feat(m7.3): MCP tenant-scoped bearer tokens

LLM clients (Claude Desktop, Cursor, ChatGPT) can't run a Keycloak
OIDC flow, so the MCP server can't use JWTs for auth. This PR
introduces opaque static bearer tokens minted per-tenant via new
agent endpoints, validated by the MCP server, and used to route
incoming MCP requests to the caller's per-tenant database.

Until now, the MCP server connected to a single shared MongoDB DB
with no auth and no tenant awareness — every tool (list_findings,
list_sbom_packages, etc.) returned data across all tenants. After
M7.2 made the agent per-tenant, MCP was the lone cross-tenant data
leak. This closes it.

Design summary
- Token format: `mcpt_<43 url-safe random chars>` (48 chars total).
  Opaque, never embeds tenant_id, never stored in plaintext.
- Storage: cross-tenant `<prefix>__admin.mcp_tokens` collection,
  keyed by SHA-256 hash. Each row carries the tenant_id, name,
  created_by, created_at, last_used_at, revoked flag.
- Agent endpoints (tenant-scoped via TenantCtx):
    POST   /api/v1/mcp-tokens    → mint (returns raw token ONCE)
    GET    /api/v1/mcp-tokens    → list (metadata + 12-char prefix,
                                   never the hash)
    DELETE /api/v1/mcp-tokens/id → soft revoke
- MCP middleware: extract `Authorization: Bearer mcpt_...`, sniff
  the prefix, SHA-256 → lookup in admin DB → reject if missing or
  revoked. Updates last_used_at fire-and-forget so it never blocks.
  Sets `tokio::task_local!` TENANT_ID for the inner service call;
  the rmcp tool handlers read it and resolve the per-tenant DB.
- task_local is scoped via TENANT_ID.scope(...) around next.run(req)
  so the rmcp tool handlers downstream see the tenant_id without
  modifying their (macro-generated) signatures.

Files
- compliance-core/src/models/mcp_token.rs (new) — McpToken +
  McpTokenView (public projection without the hash).
- compliance-agent/src/database.rs — DatabasePool::admin_db() +
  admin_db_name(): cross-tenant access for token storage.
- compliance-agent/src/api/handlers/mcp_tokens.rs (new) — three
  endpoints. Token generation: 32 random bytes → URL-safe base64,
  no padding. SHA-256 hex stored.
- compliance-mcp/src/database.rs — replaced single Database with
  DatabasePool. Tenant-scoped Database constructed per request.
  Same sanitization + 63-byte cap + hash fallback as the agent.
- compliance-mcp/src/auth.rs (new) — bearer middleware + task_local.
  Includes a SHA-256 round-trip test against a known vector.
- compliance-mcp/src/main.rs — HTTP transport: bearer middleware
  layered on /mcp (not /health, so orca's container probe still
  works). stdio transport: falls back to STDIO_TENANT_ID env (defaults
  to "dev") so local development still works; logged loudly as
  not-for-production.
- compliance-mcp/src/server.rs — each of the 12 tool handlers
  resolves the per-tenant DB via task_local before calling its tool
  fn. Tool fns themselves are unchanged.

Token UX
- Generated by the dashboard (or curl + KC JWT) — user sees raw
  token exactly once, copies it into their LLM client config.
- Dashboard UI for management is a follow-up; can use curl in the
  meantime:
    curl -X POST https://comp-dev.../api/v1/mcp-tokens \
      -H "Authorization: Bearer $KC_JWT" \
      -H "Content-Type: application/json" \
      -d '{"name":"Claude Desktop"}'

Test plan
- cargo fmt --all clean
- cargo clippy --workspace --exclude compliance-dashboard
  -- -D warnings clean
- cargo test -p compliance-core --lib — 7 pass
- cargo test -p compliance-agent --lib — 230 pass (+2 new for
  token generation + sha256 stability)
- cargo test -p compliance-agent --test tenant_isolation — 6 pass
- cargo test -p compliance-mcp — 34 pass (+1 new sha256 vector)

What's deferred
- Dashboard UI for managing tokens (page + create modal + list/
  revoke). Trivial once the API is live.
- Token expiry + per-tool scope (today every token grants access
  to all 12 tools for its tenant).
- Lifting DatabasePool into compliance-core (duplicated for now
  in compliance-mcp to keep this PR focused; lift if a third
  consumer appears).

Production
- The `<prefix>__admin` DB needs to NOT collide with a tenant
  DB. Sanitized tenant_id never starts with `_admin` for any
  current tenant_id shape (UUIDs); flagged in the database.rs
  docstring so tenant provisioning can reject `_admin*` ids
  proactively.
- orca-infra MCP service block already has MONGODB_URI /
  MONGODB_DATABASE — no new env needed. No KC creds since MCP
  doesn't use Keycloak for its own auth.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

.cargo/audit.toml

@@ -0,0 +1,10 @@
+[advisories]
+ignore = [
+    # hickory-proto 0.25.x pulled in transitively via mongodb → hickory-resolver.
+    # MongoDB 3.x has not yet released with hickory-resolver 0.26.x, so we cannot
+    # upgrade past this without a mongodb release. Both are DNS-layer DoS vectors
+    # requiring a MITM/controlled DNS server against MongoDB's hostname resolution —
+    # not a realistic attack surface here. Revisit when mongodb bumps hickory.
+    "RUSTSEC-2026-0118", # NSEC3 loop, no fix available upstream
+    "RUSTSEC-2026-0119", # O(n²) name compression, fixed in hickory-proto >=0.26.1
+]

.gitea/workflows/ci.yml

@@ -70,7 +70,7 @@ jobs:
 
       # Tests (reuses compilation artifacts from clippy)
       - name: Tests (core + agent)
-        run: cargo test -p compliance-core -p compliance-agent
+        run: cargo test -p compliance-core -p compliance-agent --lib
       - name: Tests (dashboard server)
         run: cargo test -p compliance-dashboard --features server --no-default-features
       - name: Tests (dashboard web)
@@ -145,13 +145,20 @@ jobs:
     needs: [detect-changes]
     if: needs.detect-changes.outputs.agent == 'true'
     container:
-      image: alpine:latest
+      image: docker:27-cli
     steps:
-      - name: Trigger Coolify deploy
+      - name: Build, push and trigger orca redeploy
         run: |
-          apk add --no-cache curl
-          curl -sf "${{ secrets.COOLIFY_WEBHOOK_AGENT }}" \
-            -H "Authorization: Bearer ${{ secrets.COOLIFY_TOKEN }}"
+          apk add --no-cache git curl openssl
+          git init && git remote add origin "${GITHUB_SERVER_URL}/${GITHUB_REPOSITORY}.git"
+          git fetch --depth=1 origin "${GITHUB_SHA}" && git checkout FETCH_HEAD
+          IMAGE=registry.meghsakha.com/compliance-agent
+          echo "${{ secrets.REGISTRY_PASSWORD }}" | docker login registry.meghsakha.com -u "${{ secrets.REGISTRY_USERNAME }}" --password-stdin
+          docker build -f Dockerfile.agent -t "$IMAGE:latest" -t "$IMAGE:${GITHUB_SHA}" .
+          docker push "$IMAGE:latest" && docker push "$IMAGE:${GITHUB_SHA}"
+          PAYLOAD=$(printf '{"ref":"refs/heads/main","repository":{"full_name":"sharang/compliance-scanner-agent"},"head_commit":{"id":"%s","message":"deploy agent"}}' "${GITHUB_SHA}")
+          SIG=$(printf '%s' "$PAYLOAD" | openssl dgst -sha256 -hmac "${{ secrets.ORCA_WEBHOOK_SECRET }}" | awk '{print $2}')
+          RESP=$(curl -fsS -w "\nHTTP %{http_code}" -X POST "http://46.225.100.82:6880/api/v1/webhooks/github" -H "Content-Type: application/json" -H "X-Hub-Signature-256: sha256=$SIG" -d "$PAYLOAD"); echo "$RESP"
 
   deploy-dashboard:
     name: Deploy Dashboard
@@ -159,13 +166,20 @@ jobs:
     needs: [detect-changes]
     if: needs.detect-changes.outputs.dashboard == 'true'
     container:
-      image: alpine:latest
+      image: docker:27-cli
     steps:
-      - name: Trigger Coolify deploy
+      - name: Build, push and trigger orca redeploy
         run: |
-          apk add --no-cache curl
-          curl -sf "${{ secrets.COOLIFY_WEBHOOK_DASHBOARD }}" \
-            -H "Authorization: Bearer ${{ secrets.COOLIFY_TOKEN }}"
+          apk add --no-cache git curl openssl
+          git init && git remote add origin "${GITHUB_SERVER_URL}/${GITHUB_REPOSITORY}.git"
+          git fetch --depth=1 origin "${GITHUB_SHA}" && git checkout FETCH_HEAD
+          IMAGE=registry.meghsakha.com/compliance-dashboard
+          echo "${{ secrets.REGISTRY_PASSWORD }}" | docker login registry.meghsakha.com -u "${{ secrets.REGISTRY_USERNAME }}" --password-stdin
+          docker build -f Dockerfile.dashboard -t "$IMAGE:latest" -t "$IMAGE:${GITHUB_SHA}" .
+          docker push "$IMAGE:latest" && docker push "$IMAGE:${GITHUB_SHA}"
+          PAYLOAD=$(printf '{"ref":"refs/heads/main","repository":{"full_name":"sharang/compliance-scanner-agent"},"head_commit":{"id":"%s","message":"deploy dashboard"}}' "${GITHUB_SHA}")
+          SIG=$(printf '%s' "$PAYLOAD" | openssl dgst -sha256 -hmac "${{ secrets.ORCA_WEBHOOK_SECRET }}" | awk '{print $2}')
+          RESP=$(curl -fsS -w "\nHTTP %{http_code}" -X POST "http://46.225.100.82:6880/api/v1/webhooks/github" -H "Content-Type: application/json" -H "X-Hub-Signature-256: sha256=$SIG" -d "$PAYLOAD"); echo "$RESP"
 
   deploy-docs:
     name: Deploy Docs
@@ -173,13 +187,20 @@ jobs:
     needs: [detect-changes]
     if: needs.detect-changes.outputs.docs == 'true'
     container:
-      image: alpine:latest
+      image: docker:27-cli
     steps:
-      - name: Trigger Coolify deploy
+      - name: Build, push and trigger orca redeploy
         run: |
-          apk add --no-cache curl
-          curl -sf "${{ secrets.COOLIFY_WEBHOOK_DOCS }}" \
-            -H "Authorization: Bearer ${{ secrets.COOLIFY_TOKEN }}"
+          apk add --no-cache git curl openssl
+          git init && git remote add origin "${GITHUB_SERVER_URL}/${GITHUB_REPOSITORY}.git"
+          git fetch --depth=1 origin "${GITHUB_SHA}" && git checkout FETCH_HEAD
+          IMAGE=registry.meghsakha.com/compliance-docs
+          echo "${{ secrets.REGISTRY_PASSWORD }}" | docker login registry.meghsakha.com -u "${{ secrets.REGISTRY_USERNAME }}" --password-stdin
+          docker build -f Dockerfile.docs -t "$IMAGE:latest" -t "$IMAGE:${GITHUB_SHA}" .
+          docker push "$IMAGE:latest" && docker push "$IMAGE:${GITHUB_SHA}"
+          PAYLOAD=$(printf '{"ref":"refs/heads/main","repository":{"full_name":"sharang/compliance-scanner-agent"},"head_commit":{"id":"%s","message":"deploy docs"}}' "${GITHUB_SHA}")
+          SIG=$(printf '%s' "$PAYLOAD" | openssl dgst -sha256 -hmac "${{ secrets.ORCA_WEBHOOK_SECRET }}" | awk '{print $2}')
+          RESP=$(curl -fsS -w "\nHTTP %{http_code}" -X POST "http://46.225.100.82:6880/api/v1/webhooks/github" -H "Content-Type: application/json" -H "X-Hub-Signature-256: sha256=$SIG" -d "$PAYLOAD"); echo "$RESP"
 
   deploy-mcp:
     name: Deploy MCP
@@ -187,10 +208,17 @@ jobs:
     needs: [detect-changes]
     if: needs.detect-changes.outputs.mcp == 'true'
     container:
-      image: alpine:latest
+      image: docker:27-cli
     steps:
-      - name: Trigger Coolify deploy
+      - name: Build, push and trigger orca redeploy
         run: |
-          apk add --no-cache curl
-          curl -sf "${{ secrets.COOLIFY_WEBHOOK_MCP }}" \
-            -H "Authorization: Bearer ${{ secrets.COOLIFY_TOKEN }}"
+          apk add --no-cache git curl openssl
+          git init && git remote add origin "${GITHUB_SERVER_URL}/${GITHUB_REPOSITORY}.git"
+          git fetch --depth=1 origin "${GITHUB_SHA}" && git checkout FETCH_HEAD
+          IMAGE=registry.meghsakha.com/compliance-mcp
+          echo "${{ secrets.REGISTRY_PASSWORD }}" | docker login registry.meghsakha.com -u "${{ secrets.REGISTRY_USERNAME }}" --password-stdin
+          docker build -f Dockerfile.mcp -t "$IMAGE:latest" -t "$IMAGE:${GITHUB_SHA}" .
+          docker push "$IMAGE:latest" && docker push "$IMAGE:${GITHUB_SHA}"
+          PAYLOAD=$(printf '{"ref":"refs/heads/main","repository":{"full_name":"sharang/compliance-scanner-agent"},"head_commit":{"id":"%s","message":"deploy mcp"}}' "${GITHUB_SHA}")
+          SIG=$(printf '%s' "$PAYLOAD" | openssl dgst -sha256 -hmac "${{ secrets.ORCA_WEBHOOK_SECRET }}" | awk '{print $2}')
+          RESP=$(curl -fsS -w "\nHTTP %{http_code}" -X POST "http://46.225.100.82:6880/api/v1/webhooks/github" -H "Content-Type: application/json" -H "X-Hub-Signature-256: sha256=$SIG" -d "$PAYLOAD"); echo "$RESP"

.gitea/workflows/nightly.yml

@@ -0,0 +1,52 @@
+name: Nightly E2E Tests
+
+on:
+  schedule:
+    - cron: '0 3 * * *'  # 3 AM UTC daily
+  workflow_dispatch:       # Allow manual trigger
+
+env:
+  CARGO_TERM_COLOR: always
+  RUSTFLAGS: "-D warnings"
+  RUSTC_WRAPPER: /usr/local/bin/sccache
+  SCCACHE_DIR: /tmp/sccache
+  TEST_MONGODB_URI: "mongodb://root:example@mongo:27017/?authSource=admin"
+
+concurrency:
+  group: nightly-e2e
+  cancel-in-progress: true
+
+jobs:
+  e2e:
+    name: E2E Tests
+    runs-on: docker
+    container:
+      image: rust:1.94-bookworm
+    services:
+      mongo:
+        image: mongo:7
+        env:
+          MONGO_INITDB_ROOT_USERNAME: root
+          MONGO_INITDB_ROOT_PASSWORD: example
+    steps:
+      - name: Checkout
+        run: |
+          git init
+          git remote add origin "${GITHUB_SERVER_URL}/${GITHUB_REPOSITORY}.git"
+          git fetch --depth=1 origin "${GITHUB_SHA:-refs/heads/main}"
+          git checkout FETCH_HEAD
+
+      - name: Install sccache
+        run: |
+          curl -fsSL https://github.com/mozilla/sccache/releases/download/v0.9.1/sccache-v0.9.1-x86_64-unknown-linux-musl.tar.gz \
+            | tar xz --strip-components=1 -C /usr/local/bin/ sccache-v0.9.1-x86_64-unknown-linux-musl/sccache
+          chmod +x /usr/local/bin/sccache
+        env:
+          RUSTC_WRAPPER: ""
+
+      - name: Run E2E tests
+        run: cargo test -p compliance-agent --test e2e -- --test-threads=4
+
+      - name: Show sccache stats
+        run: sccache --show-stats
+        if: always()

Cargo.lock

@@ -676,6 +676,7 @@ dependencies = [
  "jsonwebtoken",
  "mongodb",
  "octocrab",
+ "rand 0.9.2",
  "regex",
  "reqwest",
  "secrecy",
@@ -687,6 +688,7 @@ dependencies = [
  "tokio-cron-scheduler",
  "tokio-stream",
  "tokio-tungstenite 0.26.2",
+ "tower",
  "tower-http",
  "tracing",
  "tracing-subscriber",
@@ -700,19 +702,23 @@ dependencies = [
 name = "compliance-core"
 version = "0.1.0"
 dependencies = [
+ "axum",
  "bson",
  "chrono",
  "hex",
+ "jsonwebtoken",
  "mongodb",
  "opentelemetry",
  "opentelemetry-appender-tracing",
  "opentelemetry-otlp",
  "opentelemetry_sdk",
+ "reqwest",
  "secrecy",
  "serde",
  "serde_json",
  "sha2",
  "thiserror 2.0.18",
+ "tokio",
  "tracing",
  "tracing-opentelemetry",
  "tracing-subscriber",
@@ -813,12 +819,15 @@ dependencies = [
  "bson",
  "chrono",
  "compliance-core",
+ "dashmap",
  "dotenvy",
+ "hex",
  "mongodb",
  "rmcp",
  "schemars 1.2.1",
  "serde",
  "serde_json",
+ "sha2",
  "thiserror 2.0.18",
  "tokio",
  "tower-http",
@@ -826,6 +835,20 @@ dependencies = [
  "tracing-subscriber",
 ]
 
+[[package]]
+name = "compliance-smoke"
+version = "0.1.0"
+dependencies = [
+ "axum",
+ "compliance-core",
+ "reqwest",
+ "serde",
+ "serde_json",
+ "tokio",
+ "tracing",
+ "tracing-subscriber",
+]
+
 [[package]]
 name = "console_error_panic_hook"
 version = "0.1.7"
@@ -3524,9 +3547,9 @@ checksum = "224484c5d09285a7b8cb0a0c117e847ebd14cb6e4470ecf68cdb89c503b0edb9"
 
 [[package]]
 name = "mongodb"
-version = "3.5.1"
+version = "3.6.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "803dd859e8afa084c255a8effd8000ff86f7c8076a50cd6d8c99e8f3496f75c2"
+checksum = "1ef2c933617431ad0246fb5b43c425ebdae18c7f7259c87de0726d93b0e7e91b"
 dependencies = [
  "base64",
  "bitflags",
@@ -3570,9 +3593,9 @@ dependencies = [
 
 [[package]]
 name = "mongodb-internal-macros"
-version = "3.5.1"
+version = "3.6.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a973ef3dd3dbc6f6e65bbdecfd9ec5e781b9e7493b0f369a7c62e35d8e5ae2c8"
+checksum = "9e5758dc828eb2d02ec30563cba365609d56ddd833190b192beaee2b475a7bb3"
 dependencies = [
  "macro_magic",
  "proc-macro2",
@@ -4699,9 +4722,9 @@ dependencies = [
 
 [[package]]
 name = "rustls-webpki"
-version = "0.103.9"
+version = "0.103.13"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d7df23109aa6c1567d1c575b9952556388da57401e4ace1d15f79eedad0d8f53"
+checksum = "61c429a8649f110dddef65e2a5ad240f747e85f7758a6bccc7e5777bd33f756e"
 dependencies = [
  "ring",
  "rustls-pki-types",
@@ -5171,7 +5194,7 @@ version = "0.8.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c1c97747dbf44bb1ca44a561ece23508e99cb592e862f22222dcf42f51d1e451"
 dependencies = [
- "heck 0.5.0",
+ "heck 0.4.1",
  "proc-macro2",
  "quote",
  "syn",

Cargo.toml

@@ -6,6 +6,7 @@ members = [
     "compliance-graph",
     "compliance-dast",
     "compliance-mcp",
+    "compliance-smoke",
 ]
 resolver = "2"
 
@@ -33,3 +34,5 @@ zip = { version = "2", features = ["aes-crypto", "deflate"] }
 dashmap = "6"
 tokio-stream = { version = "0.1", features = ["sync"] }
 aes-gcm = "0.10"
+rand = "0.9"
+base64 = "0.22"

Dockerfile.agent

@@ -33,9 +33,15 @@ RUN pip3 install --break-system-packages ruff
 
 COPY --from=builder /app/target/release/compliance-agent /usr/local/bin/compliance-agent
 
+# Copy documentation for the help chat assistant
+COPY --from=builder /app/README.md /app/README.md
+COPY --from=builder /app/docs /app/docs
+ENV HELP_DOCS_PATH=/app
+
 # Ensure SSH key directory exists
 RUN mkdir -p /data/compliance-scanner/ssh
 
 EXPOSE 3001 3002
 
 ENTRYPOINT ["compliance-agent"]
+

Dockerfile.dashboard

@@ -1,6 +1,6 @@
 FROM rust:1.94-bookworm AS builder
 
-RUN cargo install dioxus-cli --version 0.7.3
+RUN cargo install dioxus-cli --version 0.7.3 --locked
 
 ARG DOCS_URL=/docs
 
@@ -20,3 +20,4 @@ ENV IP=0.0.0.0
 EXPOSE 8080
 
 ENTRYPOINT ["./compliance-dashboard"]
+

Dockerfile.docs

@@ -12,3 +12,4 @@ RUN rm /etc/nginx/conf.d/default.conf
 COPY docs/nginx.conf /etc/nginx/conf.d/default.conf
 COPY --from=builder /app/.vitepress/dist /usr/share/nginx/html
 EXPOSE 80
+

Dockerfile.mcp

@@ -14,3 +14,4 @@ EXPOSE 8090
 ENV MCP_PORT=8090
 
 ENTRYPOINT ["compliance-mcp"]
+

README.md

@@ -28,9 +28,9 @@
 
 ## About
 
-Compliance Scanner is an autonomous agent that continuously monitors git repositories for security vulnerabilities, GDPR/OAuth compliance patterns, and dependency risks. It creates issues in external trackers (GitHub/GitLab/Jira) with evidence and remediation suggestions, reviews pull requests, and exposes a Dioxus-based dashboard for visualization.
+Compliance Scanner is an autonomous agent that continuously monitors git repositories for security vulnerabilities, GDPR/OAuth compliance patterns, and dependency risks. It creates issues in external trackers (GitHub/GitLab/Jira/Gitea) with evidence and remediation suggestions, reviews pull requests with multi-pass LLM analysis, runs autonomous penetration tests, and exposes a Dioxus-based dashboard for visualization.
 
-> **How it works:** The agent runs as a lazy daemon -- it only scans when new commits are detected, triggered by cron schedules or webhooks. LLM-powered triage filters out false positives and generates actionable remediation.
+> **How it works:** The agent runs as a lazy daemon -- it only scans when new commits are detected, triggered by cron schedules or webhooks. LLM-powered triage filters out false positives and generates actionable remediation with multi-language awareness.
 
 ## Features
 
@@ -41,31 +41,38 @@ Compliance Scanner is an autonomous agent that continuously monitors git reposit
 | **CVE Monitoring** | OSV.dev batch queries, NVD CVSS enrichment, SearXNG context |
 | **GDPR Patterns** | Detect PII logging, missing consent, hardcoded retention, missing deletion |
 | **OAuth Patterns** | Detect implicit grant, missing PKCE, token in localStorage, token in URLs |
-| **LLM Triage** | Confidence scoring via LiteLLM to filter false positives |
-| **Issue Creation** | Auto-create issues in GitHub, GitLab, or Jira with code evidence |
-| **PR Reviews** | Post security review comments on pull requests |
-| **Dashboard** | Fullstack Dioxus UI with findings, SBOM, issues, and statistics |
-| **Webhooks** | GitHub (HMAC-SHA256) and GitLab webhook receivers for push/PR events |
+| **LLM Triage** | Multi-language-aware confidence scoring (Rust, Python, Go, Java, Ruby, PHP, C++) |
+| **Issue Creation** | Auto-create issues in GitHub, GitLab, Jira, or Gitea with dedup via fingerprints |
+| **PR Reviews** | Multi-pass security review (logic, security, convention, complexity) with dedup |
+| **DAST Scanning** | Black-box security testing with endpoint discovery and parameter fuzzing |
+| **AI Pentesting** | Autonomous LLM-orchestrated penetration testing with encrypted reports |
+| **Code Graph** | Interactive code knowledge graph with impact analysis |
+| **AI Chat (RAG)** | Natural language Q&A grounded in repository source code |
+| **Help Assistant** | Documentation-grounded help chat accessible from every dashboard page |
+| **MCP Server** | Expose live security data to Claude, Cursor, and other AI tools |
+| **Dashboard** | Fullstack Dioxus UI with findings, SBOM, issues, DAST, pentest, and graph |
+| **Webhooks** | GitHub, GitLab, and Gitea webhook receivers for push/PR events |
+| **Finding Dedup** | SHA-256 fingerprint dedup for SAST, CWE-based dedup for DAST findings |
 
 ## Architecture
 
 ```
-┌─────────────────────────────────────────────────────────────┐
-│                    Cargo Workspace                           │
-├──────────────┬──────────────────┬───────────────────────────┤
-│ compliance-  │ compliance-      │ compliance-               │
-│ core         │ agent            │ dashboard                 │
-│ (lib)        │ (bin)            │ (bin, Dioxus 0.7.3)       │
-│              │                  │                           │
-│ Models       │ Scan Pipeline    │ Fullstack Web UI          │
-│ Traits       │ LLM Client      │ Server Functions           │
-│ Config       │ Issue Trackers   │ Charts + Tables           │
-│ Errors       │ Scheduler        │ Settings Page             │
-│              │ REST API         │                           │
-│              │ Webhooks         │                           │
-└──────────────┴──────────────────┴───────────────────────────┘
-                        │
-                   MongoDB (shared)
+┌──────────────────────────────────────────────────────────────────────────┐
+│                          Cargo Workspace                                 │
+├──────────────┬──────────────────┬──────────────┬──────────┬─────────────┤
+│ compliance-  │ compliance-      │ compliance-  │ complian-│ compliance- │
+│ core (lib)   │ agent (bin)      │ dashboard    │ ce-graph │ mcp (bin)   │
+│              │                  │ (bin)        │ (lib)    │             │
+│ Models       │ Scan Pipeline    │ Dioxus 0.7   │ Tree-    │ MCP Server  │
+│ Traits       │ LLM Client      │ Fullstack UI │ sitter   │ Live data   │
+│ Config       │ Issue Trackers   │ Help Chat    │ Graph    │ for AI      │
+│ Errors       │ Pentest Engine   │ Server Fns   │ Embedds  │ tools       │
+│              │ DAST Tools       │              │ RAG      │             │
+│              │ REST API         │              │          │             │
+│              │ Webhooks         │              │          │             │
+└──────────────┴──────────────────┴──────────────┴──────────┴─────────────┘
+                                 │
+                            MongoDB (shared)
 ```
 
 ## Scan Pipeline (7 Stages)
@@ -84,11 +91,16 @@ Compliance Scanner is an autonomous agent that continuously monitors git reposit
 |-------|-----------|
 | Shared Library | `compliance-core` -- models, traits, config |
 | Agent | Axum REST API, git2, tokio-cron-scheduler, Semgrep, Syft |
-| Dashboard | Dioxus 0.7.3 fullstack, Tailwind CSS |
+| Dashboard | Dioxus 0.7.3 fullstack, Tailwind CSS 4 |
+| Code Graph | `compliance-graph` -- tree-sitter parsing, embeddings, RAG |
+| MCP Server | `compliance-mcp` -- Model Context Protocol for AI tools |
+| DAST | `compliance-dast` -- dynamic application security testing |
 | Database | MongoDB with typed collections |
-| LLM | LiteLLM (OpenAI-compatible API) |
-| Issue Trackers | GitHub (octocrab), GitLab (REST v4), Jira (REST v3) |
+| LLM | LiteLLM (OpenAI-compatible API for chat, triage, embeddings) |
+| Issue Trackers | GitHub (octocrab), GitLab (REST v4), Jira (REST v3), Gitea |
 | CVE Sources | OSV.dev, NVD, SearXNG |
+| Auth | Keycloak (OAuth2/PKCE, SSO) |
+| Browser Automation | Chromium (headless, for pentesting and PDF generation) |
 
 ## Getting Started
 
@@ -151,20 +163,35 @@ The agent exposes a REST API on port 3001:
 | `GET` | `/api/v1/sbom` | List dependencies |
 | `GET` | `/api/v1/issues` | List cross-tracker issues |
 | `GET` | `/api/v1/scan-runs` | Scan execution history |
+| `GET` | `/api/v1/graph/:repo_id` | Code knowledge graph |
+| `POST` | `/api/v1/graph/:repo_id/build` | Trigger graph build |
+| `GET` | `/api/v1/dast/targets` | List DAST targets |
+| `POST` | `/api/v1/dast/targets` | Add DAST target |
+| `GET` | `/api/v1/dast/findings` | List DAST findings |
+| `POST` | `/api/v1/chat/:repo_id` | RAG-powered code chat |
+| `POST` | `/api/v1/help/chat` | Documentation-grounded help chat |
+| `POST` | `/api/v1/pentest/sessions` | Create pentest session |
+| `POST` | `/api/v1/pentest/sessions/:id/export` | Export encrypted pentest report |
 | `POST` | `/webhook/github` | GitHub webhook (HMAC-SHA256) |
 | `POST` | `/webhook/gitlab` | GitLab webhook (token verify) |
+| `POST` | `/webhook/gitea` | Gitea webhook |
 
 ## Dashboard Pages
 
 | Page | Description |
 |------|-------------|
-| **Overview** | Stat cards, severity distribution chart |
-| **Repositories** | Add/manage tracked repos, trigger scans |
-| **Findings** | Filterable table by severity, type, status |
+| **Overview** | Stat cards, severity distribution, AI chat cards, MCP status |
+| **Repositories** | Add/manage tracked repos, trigger scans, webhook config |
+| **Findings** | Filterable table by severity, type, status, scanner |
 | **Finding Detail** | Code evidence, remediation, suggested fix, linked issue |
-| **SBOM** | Dependency inventory with vulnerability badges |
-| **Issues** | Cross-tracker view (GitHub + GitLab + Jira) |
-| **Settings** | Configure LiteLLM, tracker tokens, SearXNG URL |
+| **SBOM** | Dependency inventory with vulnerability badges, license summary |
+| **Issues** | Cross-tracker view (GitHub + GitLab + Jira + Gitea) |
+| **Code Graph** | Interactive architecture visualization, impact analysis |
+| **AI Chat** | RAG-powered Q&A about repository code |
+| **DAST** | Dynamic scanning targets, findings, and scan history |
+| **Pentest** | AI-driven pentest sessions, attack chain visualization |
+| **MCP Servers** | Model Context Protocol server management |
+| **Help Chat** | Floating assistant (available on every page) for product Q&A |
 
 ## Project Structure
 
@@ -173,19 +200,24 @@ compliance-scanner/
 ├── compliance-core/        Shared library (models, traits, config, errors)
 ├── compliance-agent/       Agent daemon (pipeline, LLM, trackers, API, webhooks)
 │   └── src/
-│       ├── pipeline/       7-stage scan pipeline
-│       ├── llm/            LiteLLM client, triage, descriptions, fixes, PR review
-│       ├── trackers/       GitHub, GitLab, Jira integrations
-│       ├── api/            REST API (Axum)
-│       └── webhooks/       GitHub + GitLab webhook receivers
+│       ├── pipeline/       7-stage scan pipeline, dedup, PR reviews, code review
+│       ├── llm/            LiteLLM client, triage, descriptions, fixes, review prompts
+│       ├── trackers/       GitHub, GitLab, Jira, Gitea integrations
+│       ├── pentest/        AI-driven pentest orchestrator, tools, reports
+│       ├── rag/            RAG pipeline, chunking, embedding
+│       ├── api/            REST API (Axum), help chat
+│       └── webhooks/       GitHub, GitLab, Gitea webhook receivers
 ├── compliance-dashboard/   Dioxus fullstack dashboard
 │   └── src/
-│       ├── components/     Reusable UI components
-│       ├── infrastructure/ Server functions, DB, config
-│       └── pages/          Full page views
+│       ├── components/     Reusable UI (sidebar, help chat, attack chain, etc.)
+│       ├── infrastructure/ Server functions, DB, config, auth
+│       └── pages/          Full page views (overview, DAST, pentest, graph, etc.)
+├── compliance-graph/       Code knowledge graph (tree-sitter, embeddings, RAG)
+├── compliance-dast/        Dynamic application security testing
+├── compliance-mcp/         Model Context Protocol server
+├── docs/                   VitePress documentation site
 ├── assets/                 Static assets (CSS, icons)
-├── styles/                 Tailwind input stylesheet
-└── bin/                    Dashboard binary entrypoint
+└── styles/                 Tailwind input stylesheet
 ```
 
 ## External Services
@@ -193,10 +225,12 @@ compliance-scanner/
 | Service | Purpose | Default URL |
 |---------|---------|-------------|
 | MongoDB | Persistence | `mongodb://localhost:27017` |
-| LiteLLM | LLM proxy for triage and generation | `http://localhost:4000` |
+| LiteLLM | LLM proxy (chat, triage, embeddings) | `http://localhost:4000` |
 | SearXNG | CVE context search | `http://localhost:8888` |
+| Keycloak | Authentication (OAuth2/PKCE, SSO) | `http://localhost:8080` |
 | Semgrep | SAST scanning | CLI tool |
 | Syft | SBOM generation | CLI tool |
+| Chromium | Headless browser (pentesting, PDF) | Managed via Docker |
 
 ---
 

compliance-agent/Cargo.toml

@@ -7,7 +7,7 @@ edition = "2021"
 workspace = true
 
 [dependencies]
-compliance-core = { workspace = true, features = ["mongodb", "telemetry"] }
+compliance-core = { workspace = true, features = ["mongodb", "telemetry", "axum"] }
 compliance-graph = { path = "../compliance-graph" }
 compliance-dast = { path = "../compliance-dast" }
 serde = { workspace = true }
@@ -25,7 +25,7 @@ uuid = { workspace = true }
 secrecy = { workspace = true }
 regex = { workspace = true }
 axum = "0.8"
-tower-http = { version = "0.6", features = ["cors", "trace"] }
+tower-http = { version = "0.6", features = ["cors", "trace", "set-header"] }
 git2 = "0.20"
 octocrab = "0.44"
 tokio-cron-scheduler = "0.13"
@@ -42,3 +42,16 @@ tokio-tungstenite = { version = "0.26", features = ["rustls-tls-webpki-roots"] }
 futures-core = "0.3"
 dashmap = { workspace = true }
 tokio-stream = { workspace = true }
+rand = { workspace = true }
+
+[dev-dependencies]
+compliance-core = { workspace = true, features = ["mongodb", "axum"] }
+tower = { version = "0.5", features = ["util"] }
+reqwest = { workspace = true }
+serde_json = { workspace = true }
+tokio = { workspace = true }
+mongodb = { workspace = true }
+uuid = { workspace = true }
+secrecy = { workspace = true }
+axum = "0.8"
+tower-http = { version = "0.6", features = ["cors"] }

compliance-agent/src/agent.rs

@@ -6,7 +6,7 @@ use tokio::sync::{broadcast, watch, Semaphore};
 use compliance_core::models::pentest::PentestEvent;
 use compliance_core::AgentConfig;
 
-use crate::database::Database;
+use crate::database::DatabasePool;
 use crate::llm::LlmClient;
 use crate::pipeline::orchestrator::PipelineOrchestrator;
 
@@ -16,7 +16,10 @@ const DEFAULT_MAX_CONCURRENT_SESSIONS: usize = 5;
 #[derive(Clone)]
 pub struct ComplianceAgent {
     pub config: AgentConfig,
-    pub db: Database,
+    /// Per-tenant Mongo broker. Every code path must obtain a
+    /// tenant-scoped [`crate::database::Database`] from this pool —
+    /// there is no single shared database any more.
+    pub db_pool: DatabasePool,
     pub llm: Arc<LlmClient>,
     pub http: reqwest::Client,
     /// Per-session broadcast senders for SSE streaming.
@@ -28,18 +31,23 @@ pub struct ComplianceAgent {
 }
 
 impl ComplianceAgent {
-    pub fn new(config: AgentConfig, db: Database) -> Self {
+    pub fn new(config: AgentConfig, db_pool: DatabasePool) -> Self {
         let llm = Arc::new(LlmClient::new(
             config.litellm_url.clone(),
             config.litellm_api_key.clone(),
             config.litellm_model.clone(),
             config.litellm_embed_model.clone(),
         ));
+        let http = reqwest::Client::builder()
+            .timeout(std::time::Duration::from_secs(30))
+            .connect_timeout(std::time::Duration::from_secs(10))
+            .build()
+            .unwrap_or_default();
         Self {
             config,
-            db,
+            db_pool,
             llm,
-            http: reqwest::Client::new(),
+            http,
             session_streams: Arc::new(DashMap::new()),
             session_pause: Arc::new(DashMap::new()),
             session_semaphore: Arc::new(Semaphore::new(DEFAULT_MAX_CONCURRENT_SESSIONS)),
@@ -48,28 +56,27 @@ impl ComplianceAgent {
 
     pub async fn run_scan(
         &self,
+        tenant_id: &str,
         repo_id: &str,
         trigger: compliance_core::models::ScanTrigger,
     ) -> Result<(), crate::error::AgentError> {
-        let orchestrator = PipelineOrchestrator::new(
-            self.config.clone(),
-            self.db.clone(),
-            self.llm.clone(),
-            self.http.clone(),
-        );
+        let db = self.db_pool.for_tenant_id(tenant_id).await?;
+        let orchestrator =
+            PipelineOrchestrator::new(self.config.clone(), db, self.llm.clone(), self.http.clone());
         orchestrator.run(repo_id, trigger).await
     }
 
     /// Run a PR review: scan the diff and post review comments.
     pub async fn run_pr_review(
         &self,
+        tenant_id: &str,
         repo_id: &str,
         pr_number: u64,
         base_sha: &str,
         head_sha: &str,
     ) -> Result<(), crate::error::AgentError> {
-        let repo = self
-            .db
+        let db = self.db_pool.for_tenant_id(tenant_id).await?;
+        let repo = db
             .repositories()
             .find_one(mongodb::bson::doc! {
                 "_id": mongodb::bson::oid::ObjectId::parse_str(repo_id)
@@ -80,12 +87,8 @@ impl ComplianceAgent {
                 crate::error::AgentError::Other(format!("Repository {repo_id} not found"))
             })?;
 
-        let orchestrator = PipelineOrchestrator::new(
-            self.config.clone(),
-            self.db.clone(),
-            self.llm.clone(),
-            self.http.clone(),
-        );
+        let orchestrator =
+            PipelineOrchestrator::new(self.config.clone(), db, self.llm.clone(), self.http.clone());
         orchestrator
             .run_pr_review(&repo, repo_id, pr_number, base_sha, head_sha)
             .await

compliance-agent/src/api/auth_middleware.rs

@@ -1,113 +0,0 @@
-use std::sync::Arc;
-
-use axum::{
-    extract::Request,
-    middleware::Next,
-    response::{IntoResponse, Response},
-};
-use jsonwebtoken::{decode, decode_header, jwk::JwkSet, DecodingKey, Validation};
-use reqwest::StatusCode;
-use serde::Deserialize;
-use tokio::sync::RwLock;
-
-/// Cached JWKS from Keycloak for token validation.
-#[derive(Clone)]
-pub struct JwksState {
-    pub jwks: Arc<RwLock<Option<JwkSet>>>,
-    pub jwks_url: String,
-}
-
-#[derive(Debug, Deserialize)]
-struct Claims {
-    #[allow(dead_code)]
-    sub: String,
-}
-
-const PUBLIC_ENDPOINTS: &[&str] = &["/api/v1/health"];
-
-/// Middleware that validates Bearer JWT tokens against Keycloak's JWKS.
-///
-/// Skips validation for health check endpoints.
-/// If `JwksState` is not present as an extension (keycloak not configured),
-/// all requests pass through.
-pub async fn require_jwt_auth(request: Request, next: Next) -> Response {
-    let path = request.uri().path();
-
-    if PUBLIC_ENDPOINTS.contains(&path) {
-        return next.run(request).await;
-    }
-
-    let jwks_state = match request.extensions().get::<JwksState>() {
-        Some(s) => s.clone(),
-        None => return next.run(request).await,
-    };
-
-    let auth_header = match request.headers().get("authorization") {
-        Some(h) => h,
-        None => return (StatusCode::UNAUTHORIZED, "Missing authorization header").into_response(),
-    };
-
-    let token = match auth_header.to_str() {
-        Ok(s) if s.starts_with("Bearer ") => &s[7..],
-        _ => return (StatusCode::UNAUTHORIZED, "Invalid authorization header").into_response(),
-    };
-
-    match validate_token(token, &jwks_state).await {
-        Ok(()) => next.run(request).await,
-        Err(e) => {
-            tracing::warn!("JWT validation failed: {e}");
-            (StatusCode::UNAUTHORIZED, "Invalid token").into_response()
-        }
-    }
-}
-
-async fn validate_token(token: &str, state: &JwksState) -> Result<(), String> {
-    let header = decode_header(token).map_err(|e| format!("failed to decode JWT header: {e}"))?;
-
-    let kid = header
-        .kid
-        .ok_or_else(|| "JWT missing kid header".to_string())?;
-
-    let jwks = fetch_or_get_jwks(state).await?;
-
-    let jwk = jwks
-        .keys
-        .iter()
-        .find(|k| k.common.key_id.as_deref() == Some(&kid))
-        .ok_or_else(|| "no matching key found in JWKS".to_string())?;
-
-    let decoding_key =
-        DecodingKey::from_jwk(jwk).map_err(|e| format!("failed to create decoding key: {e}"))?;
-
-    let mut validation = Validation::new(header.alg);
-    validation.validate_exp = true;
-    validation.validate_aud = false;
-
-    decode::<Claims>(token, &decoding_key, &validation)
-        .map_err(|e| format!("token validation failed: {e}"))?;
-
-    Ok(())
-}
-
-async fn fetch_or_get_jwks(state: &JwksState) -> Result<JwkSet, String> {
-    {
-        let cached = state.jwks.read().await;
-        if let Some(ref jwks) = *cached {
-            return Ok(jwks.clone());
-        }
-    }
-
-    let resp = reqwest::get(&state.jwks_url)
-        .await
-        .map_err(|e| format!("failed to fetch JWKS: {e}"))?;
-
-    let jwks: JwkSet = resp
-        .json()
-        .await
-        .map_err(|e| format!("failed to parse JWKS: {e}"))?;
-
-    let mut cached = state.jwks.write().await;
-    *cached = Some(jwks.clone());
-
-    Ok(jwks)
-}

compliance-agent/src/api/handlers/chat.rs

@@ -7,11 +7,13 @@ use mongodb::bson::doc;
 
 use compliance_core::models::chat::{ChatRequest, ChatResponse, SourceReference};
 use compliance_core::models::embedding::EmbeddingBuildRun;
+use compliance_core::tenant_ctx::TenantCtx;
 use compliance_graph::graph::embedding_store::EmbeddingStore;
 
 use crate::agent::ComplianceAgent;
 use crate::rag::pipeline::RagPipeline;
 
+use super::dto::tenant_db;
 use super::ApiResponse;
 
 type AgentExt = Extension<Arc<ComplianceAgent>>;
@@ -20,10 +22,12 @@ type AgentExt = Extension<Arc<ComplianceAgent>>;
 #[tracing::instrument(skip_all, fields(repo_id = %repo_id))]
 pub async fn chat(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(repo_id): Path<String>,
     Json(req): Json<ChatRequest>,
 ) -> Result<Json<ApiResponse<ChatResponse>>, StatusCode> {
-    let pipeline = RagPipeline::new(agent.llm.clone(), agent.db.inner());
+    let db = tenant_db(&agent, &tenant).await?;
+    let pipeline = RagPipeline::new(agent.llm.clone(), db.inner());
 
     // Step 1: Embed the user's message
     let query_vectors = agent
@@ -90,10 +94,13 @@ pub async fn chat(
     };
 
     let system_prompt = format!(
-        "You are an expert code assistant for a software repository. \
-         Answer the user's question based on the code context below. \
-         Reference specific files and functions when relevant. \
-         If the context doesn't contain enough information, say so.\n\n\
+        "You are a code assistant for this repository. Answer questions using the code context below.\n\n\
+         Rules:\n\
+         - Reference specific files, functions, and line numbers\n\
+         - Show code snippets when they help explain the answer\n\
+         - If the context is insufficient, say what's missing rather than guessing\n\
+         - Be concise — lead with the answer, then explain if needed\n\
+         - For security questions, note relevant CWEs and link to the finding if one exists\n\n\
          ## Code Context\n\n{code_context}"
     );
 
@@ -130,12 +137,15 @@ pub async fn chat(
 #[tracing::instrument(skip_all, fields(repo_id = %repo_id))]
 pub async fn build_embeddings(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(repo_id): Path<String>,
 ) -> Result<Json<serde_json::Value>, StatusCode> {
+    // Resolve the tenant DB up front so we can move it into the spawn;
+    // the JWT/dev context isn't available inside detached tasks.
+    let db = tenant_db(&agent, &tenant).await?;
     let agent_clone = (*agent).clone();
     tokio::spawn(async move {
-        let repo = match agent_clone
-            .db
+        let repo = match db
             .repositories()
             .find_one(doc! { "_id": mongodb::bson::oid::ObjectId::parse_str(&repo_id).ok() })
             .await
@@ -148,8 +158,7 @@ pub async fn build_embeddings(
         };
 
         // Get latest graph build
-        let build = match agent_clone
-            .db
+        let build = match db
             .graph_builds()
             .find_one(doc! { "repo_id": &repo_id })
             .sort(doc! { "started_at": -1 })
@@ -168,26 +177,22 @@ pub async fn build_embeddings(
             .unwrap_or_else(|| "unknown".to_string());
 
         // Get nodes
-        let nodes: Vec<compliance_core::models::graph::CodeNode> = match agent_clone
-            .db
-            .graph_nodes()
-            .find(doc! { "repo_id": &repo_id })
-            .await
-        {
-            Ok(cursor) => {
-                use futures_util::StreamExt;
-                let mut items = Vec::new();
-                let mut cursor = cursor;
-                while let Some(Ok(item)) = cursor.next().await {
-                    items.push(item);
+        let nodes: Vec<compliance_core::models::graph::CodeNode> =
+            match db.graph_nodes().find(doc! { "repo_id": &repo_id }).await {
+                Ok(cursor) => {
+                    use futures_util::StreamExt;
+                    let mut items = Vec::new();
+                    let mut cursor = cursor;
+                    while let Some(Ok(item)) = cursor.next().await {
+                        items.push(item);
+                    }
+                    items
                 }
-                items
-            }
-            Err(e) => {
-                tracing::error!("[{repo_id}] Failed to fetch nodes: {e}");
-                return;
-            }
-        };
+                Err(e) => {
+                    tracing::error!("[{repo_id}] Failed to fetch nodes: {e}");
+                    return;
+                }
+            };
 
         let creds = crate::pipeline::git::RepoCredentials {
             ssh_key_path: Some(agent_clone.config.ssh_key_path.clone()),
@@ -204,7 +209,7 @@ pub async fn build_embeddings(
             }
         };
 
-        let pipeline = RagPipeline::new(agent_clone.llm.clone(), agent_clone.db.inner());
+        let pipeline = RagPipeline::new(agent_clone.llm.clone(), db.inner());
         match pipeline
             .build_embeddings(&repo_id, &repo_path, &graph_build_id, &nodes)
             .await
@@ -231,9 +236,11 @@ pub async fn build_embeddings(
 #[tracing::instrument(skip_all, fields(repo_id = %repo_id))]
 pub async fn embedding_status(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(repo_id): Path<String>,
 ) -> Result<Json<ApiResponse<Option<EmbeddingBuildRun>>>, StatusCode> {
-    let store = EmbeddingStore::new(agent.db.inner());
+    let db = tenant_db(&agent, &tenant).await?;
+    let store = EmbeddingStore::new(db.inner());
     let build = store.get_latest_build(&repo_id).await.map_err(|e| {
         tracing::error!("Failed to get embedding status: {e}");
         StatusCode::INTERNAL_SERVER_ERROR

compliance-agent/src/api/handlers/dast.rs

@@ -7,9 +7,11 @@ use mongodb::bson::doc;
 use serde::Deserialize;
 
 use compliance_core::models::dast::{DastFinding, DastScanRun, DastTarget, DastTargetType};
+use compliance_core::tenant_ctx::TenantCtx;
 
 use crate::agent::ComplianceAgent;
 
+use super::dto::tenant_db;
 use super::{collect_cursor_async, ApiResponse, PaginationParams};
 
 type AgentExt = Extension<Arc<ComplianceAgent>>;
@@ -45,9 +47,11 @@ fn default_rate_limit() -> u32 {
 #[tracing::instrument(skip_all)]
 pub async fn list_targets(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Query(params): Query<PaginationParams>,
 ) -> Result<Json<ApiResponse<Vec<DastTarget>>>, StatusCode> {
-    let db = &agent.db;
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
     let skip = (params.page.saturating_sub(1)) * params.limit as u64;
     let total = db
         .dast_targets()
@@ -80,6 +84,7 @@ pub async fn list_targets(
 #[tracing::instrument(skip_all)]
 pub async fn add_target(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Json(req): Json<AddTargetRequest>,
 ) -> Result<Json<ApiResponse<DastTarget>>, StatusCode> {
     let mut target = DastTarget::new(req.name, req.base_url, req.target_type);
@@ -89,9 +94,8 @@ pub async fn add_target(
     target.rate_limit = req.rate_limit;
     target.allow_destructive = req.allow_destructive;
 
-    agent
-        .db
-        .dast_targets()
+    let db = tenant_db(&agent, &tenant).await?;
+    db.dast_targets()
         .insert_one(&target)
         .await
         .map_err(|_| StatusCode::INTERNAL_SERVER_ERROR)?;
@@ -107,19 +111,19 @@ pub async fn add_target(
 #[tracing::instrument(skip_all, fields(target_id = %id))]
 pub async fn trigger_scan(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(id): Path<String>,
 ) -> Result<Json<serde_json::Value>, StatusCode> {
     let oid = mongodb::bson::oid::ObjectId::parse_str(&id).map_err(|_| StatusCode::BAD_REQUEST)?;
+    let db = tenant_db(&agent, &tenant).await?;
 
-    let target = agent
-        .db
+    let target = db
         .dast_targets()
         .find_one(doc! { "_id": oid })
         .await
         .map_err(|_| StatusCode::INTERNAL_SERVER_ERROR)?
         .ok_or(StatusCode::NOT_FOUND)?;
 
-    let db = agent.db.clone();
     tokio::spawn(async move {
         let orchestrator = compliance_dast::DastOrchestrator::new(100);
         match orchestrator.run_scan(&target, Vec::new()).await {
@@ -147,9 +151,11 @@ pub async fn trigger_scan(
 #[tracing::instrument(skip_all)]
 pub async fn list_scan_runs(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Query(params): Query<PaginationParams>,
 ) -> Result<Json<ApiResponse<Vec<DastScanRun>>>, StatusCode> {
-    let db = &agent.db;
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
     let skip = (params.page.saturating_sub(1)) * params.limit as u64;
     let total = db
         .dast_scan_runs()
@@ -183,9 +189,11 @@ pub async fn list_scan_runs(
 #[tracing::instrument(skip_all)]
 pub async fn list_findings(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Query(params): Query<PaginationParams>,
 ) -> Result<Json<ApiResponse<Vec<DastFinding>>>, StatusCode> {
-    let db = &agent.db;
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
     let skip = (params.page.saturating_sub(1)) * params.limit as u64;
     let total = db
         .dast_findings()
@@ -219,12 +227,13 @@ pub async fn list_findings(
 #[tracing::instrument(skip_all, fields(finding_id = %id))]
 pub async fn get_finding(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(id): Path<String>,
 ) -> Result<Json<ApiResponse<DastFinding>>, StatusCode> {
     let oid = mongodb::bson::oid::ObjectId::parse_str(&id).map_err(|_| StatusCode::BAD_REQUEST)?;
+    let db = tenant_db(&agent, &tenant).await?;
 
-    let finding = agent
-        .db
+    let finding = db
         .dast_findings()
         .find_one(doc! { "_id": oid })
         .await

compliance-agent/src/api/handlers/dto.rs

@@ -180,6 +180,27 @@ pub struct SbomVersionDiff {
 pub(crate) type AgentExt = axum::extract::Extension<std::sync::Arc<crate::agent::ComplianceAgent>>;
 pub(crate) type ApiResult<T> = Result<axum::Json<ApiResponse<T>>, axum::http::StatusCode>;
 
+/// Resolve a tenant-scoped [`Database`] from the request's
+/// [`TenantContext`] (inserted by the M7.1 JWT middleware, or by the
+/// dev fallback in unsecured environments). The pool ensures the
+/// tenant's indexes idempotently.
+///
+/// Returns 500 on the rare path where Mongo refuses the database
+/// handle — the M7.1 auth/status middleware already rejects every
+/// other failure mode with 4xx before we get here.
+pub(crate) async fn tenant_db(
+    agent: &crate::agent::ComplianceAgent,
+    tenant: &compliance_core::tenant_ctx::TenantCtx,
+) -> Result<crate::database::Database, axum::http::StatusCode> {
+    agent.db_pool.for_tenant(&tenant.0).await.map_err(|e| {
+        tracing::error!(
+            tenant_id = %tenant.0.tenant_id,
+            "Failed to acquire tenant database: {e}"
+        );
+        axum::http::StatusCode::INTERNAL_SERVER_ERROR
+    })
+}
+
 pub(crate) async fn collect_cursor_async<T: serde::de::DeserializeOwned + Unpin + Send>(
     mut cursor: mongodb::Cursor<T>,
 ) -> Vec<T> {

compliance-agent/src/api/handlers/findings.rs

@@ -5,13 +5,16 @@ use mongodb::bson::doc;
 
 use super::dto::*;
 use compliance_core::models::Finding;
+use compliance_core::tenant_ctx::TenantCtx;
 
 #[tracing::instrument(skip_all, fields(repo_id = ?filter.repo_id, severity = ?filter.severity, scan_type = ?filter.scan_type))]
 pub async fn list_findings(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Query(filter): Query<FindingsFilter>,
 ) -> ApiResult<Vec<Finding>> {
-    let db = &agent.db;
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
     let mut query = doc! {};
     if let Some(repo_id) = &filter.repo_id {
         query.insert("repo_id", repo_id);
@@ -81,11 +84,12 @@ pub async fn list_findings(
 #[tracing::instrument(skip_all, fields(finding_id = %id))]
 pub async fn get_finding(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(id): Path<String>,
 ) -> Result<Json<ApiResponse<Finding>>, StatusCode> {
     let oid = mongodb::bson::oid::ObjectId::parse_str(&id).map_err(|_| StatusCode::BAD_REQUEST)?;
-    let finding = agent
-        .db
+    let db = tenant_db(&agent, &tenant).await?;
+    let finding = db
         .findings()
         .find_one(doc! { "_id": oid })
         .await
@@ -102,14 +106,14 @@ pub async fn get_finding(
 #[tracing::instrument(skip_all, fields(finding_id = %id))]
 pub async fn update_finding_status(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(id): Path<String>,
     Json(req): Json<UpdateStatusRequest>,
 ) -> Result<Json<serde_json::Value>, StatusCode> {
     let oid = mongodb::bson::oid::ObjectId::parse_str(&id).map_err(|_| StatusCode::BAD_REQUEST)?;
+    let db = tenant_db(&agent, &tenant).await?;
 
-    agent
-        .db
-        .findings()
+    db.findings()
         .update_one(
             doc! { "_id": oid },
             doc! { "$set": { "status": &req.status, "updated_at": mongodb::bson::DateTime::now() } },
@@ -123,6 +127,7 @@ pub async fn update_finding_status(
 #[tracing::instrument(skip_all)]
 pub async fn bulk_update_finding_status(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Json(req): Json<BulkUpdateStatusRequest>,
 ) -> Result<Json<serde_json::Value>, StatusCode> {
     let oids: Vec<mongodb::bson::oid::ObjectId> = req
@@ -135,8 +140,8 @@ pub async fn bulk_update_finding_status(
         return Err(StatusCode::BAD_REQUEST);
     }
 
-    let result = agent
-        .db
+    let db = tenant_db(&agent, &tenant).await?;
+    let result = db
         .findings()
         .update_many(
             doc! { "_id": { "$in": oids } },
@@ -153,14 +158,14 @@ pub async fn bulk_update_finding_status(
 #[tracing::instrument(skip_all)]
 pub async fn update_finding_feedback(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(id): Path<String>,
     Json(req): Json<UpdateFeedbackRequest>,
 ) -> Result<Json<serde_json::Value>, StatusCode> {
     let oid = mongodb::bson::oid::ObjectId::parse_str(&id).map_err(|_| StatusCode::BAD_REQUEST)?;
+    let db = tenant_db(&agent, &tenant).await?;
 
-    agent
-        .db
-        .findings()
+    db.findings()
         .update_one(
             doc! { "_id": oid },
             doc! { "$set": { "developer_feedback": &req.feedback, "updated_at": mongodb::bson::DateTime::now() } },

compliance-agent/src/api/handlers/graph.rs

@@ -7,9 +7,11 @@ use mongodb::bson::doc;
 use serde::{Deserialize, Serialize};
 
 use compliance_core::models::graph::{CodeEdge, CodeNode, GraphBuildRun, ImpactAnalysis};
+use compliance_core::tenant_ctx::TenantCtx;
 
 use crate::agent::ComplianceAgent;
 
+use super::dto::tenant_db;
 use super::{collect_cursor_async, ApiResponse};
 
 type AgentExt = Extension<Arc<ComplianceAgent>>;
@@ -36,9 +38,11 @@ fn default_search_limit() -> usize {
 #[tracing::instrument(skip_all, fields(repo_id = %repo_id))]
 pub async fn get_graph(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(repo_id): Path<String>,
 ) -> Result<Json<ApiResponse<GraphData>>, StatusCode> {
-    let db = &agent.db;
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
 
     // Get latest build
     let build: Option<GraphBuildRun> = db
@@ -98,9 +102,11 @@ pub async fn get_graph(
 #[tracing::instrument(skip_all, fields(repo_id = %repo_id))]
 pub async fn get_nodes(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(repo_id): Path<String>,
 ) -> Result<Json<ApiResponse<Vec<CodeNode>>>, StatusCode> {
-    let db = &agent.db;
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
     let filter = doc! { "repo_id": &repo_id };
 
     let nodes: Vec<CodeNode> = match db.graph_nodes().find(filter).await {
@@ -123,9 +129,11 @@ pub async fn get_nodes(
 #[tracing::instrument(skip_all, fields(repo_id = %repo_id))]
 pub async fn get_communities(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(repo_id): Path<String>,
 ) -> Result<Json<ApiResponse<Vec<CommunityInfo>>>, StatusCode> {
-    let db = &agent.db;
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
     let filter = doc! { "repo_id": &repo_id };
 
     let nodes: Vec<CodeNode> = match db.graph_nodes().find(filter).await {
@@ -176,9 +184,11 @@ pub struct CommunityInfo {
 #[tracing::instrument(skip_all, fields(repo_id = %repo_id, finding_id = %finding_id))]
 pub async fn get_impact(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path((repo_id, finding_id)): Path<(String, String)>,
 ) -> Result<Json<ApiResponse<Option<ImpactAnalysis>>>, StatusCode> {
-    let db = &agent.db;
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
     let filter = doc! { "repo_id": &repo_id, "finding_id": &finding_id };
 
     let impact = db
@@ -198,10 +208,12 @@ pub async fn get_impact(
 #[tracing::instrument(skip_all, fields(repo_id = %repo_id, query = %params.q))]
 pub async fn search_symbols(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(repo_id): Path<String>,
     Query(params): Query<SearchParams>,
 ) -> Result<Json<ApiResponse<Vec<CodeNode>>>, StatusCode> {
-    let db = &agent.db;
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
 
     // Simple text search on qualified_name and name fields
     let filter = doc! {
@@ -234,10 +246,12 @@ pub async fn search_symbols(
 #[tracing::instrument(skip_all, fields(repo_id = %repo_id))]
 pub async fn get_file_content(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(repo_id): Path<String>,
     Query(params): Query<FileContentParams>,
 ) -> Result<Json<ApiResponse<FileContent>>, StatusCode> {
-    let db = &agent.db;
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
 
     // Look up the repository to get repo name
     let repo = db
@@ -296,12 +310,13 @@ pub struct FileContent {
 #[tracing::instrument(skip_all, fields(repo_id = %repo_id))]
 pub async fn trigger_build(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(repo_id): Path<String>,
 ) -> Result<Json<serde_json::Value>, StatusCode> {
+    let db = tenant_db(&agent, &tenant).await?;
     let agent_clone = (*agent).clone();
     tokio::spawn(async move {
-        let repo = match agent_clone
-            .db
+        let repo = match db
             .repositories()
             .find_one(doc! { "_id": mongodb::bson::oid::ObjectId::parse_str(&repo_id).ok() })
             .await
@@ -333,8 +348,7 @@ pub async fn trigger_build(
 
         match engine.build_graph(&repo_path, &repo_id, &graph_build_id) {
             Ok((code_graph, build_run)) => {
-                let store =
-                    compliance_graph::graph::persistence::GraphStore::new(agent_clone.db.inner());
+                let store = compliance_graph::graph::persistence::GraphStore::new(db.inner());
                 let _ = store.delete_repo_graph(&repo_id).await;
                 let _ = store
                     .store_graph(&build_run, &code_graph.nodes, &code_graph.edges)

compliance-agent/src/api/handlers/health.rs

@@ -3,6 +3,7 @@ use mongodb::bson::doc;
 
 use super::dto::*;
 use compliance_core::models::ScanRun;
+use compliance_core::tenant_ctx::TenantCtx;
 
 #[tracing::instrument(skip_all)]
 pub async fn health() -> Json<serde_json::Value> {
@@ -10,8 +11,12 @@ pub async fn health() -> Json<serde_json::Value> {
 }
 
 #[tracing::instrument(skip_all)]
-pub async fn stats_overview(axum::extract::Extension(agent): AgentExt) -> ApiResult<OverviewStats> {
-    let db = &agent.db;
+pub async fn stats_overview(
+    axum::extract::Extension(agent): AgentExt,
+    tenant: TenantCtx,
+) -> ApiResult<OverviewStats> {
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
 
     let total_repositories = db
         .repositories()

compliance-agent/src/api/handlers/help_chat.rs

@@ -0,0 +1,217 @@
+use std::path::{Path, PathBuf};
+use std::sync::OnceLock;
+
+use axum::extract::Extension;
+use axum::http::StatusCode;
+use axum::Json;
+use serde::{Deserialize, Serialize};
+use walkdir::WalkDir;
+
+use super::dto::{AgentExt, ApiResponse};
+
+// ── DTOs ─────────────────────────────────────────────────────────────────────
+
+#[derive(Debug, Deserialize)]
+pub struct HelpChatMessage {
+    pub role: String,
+    pub content: String,
+}
+
+#[derive(Debug, Deserialize)]
+pub struct HelpChatRequest {
+    pub message: String,
+    #[serde(default)]
+    pub history: Vec<HelpChatMessage>,
+}
+
+#[derive(Debug, Serialize)]
+pub struct HelpChatResponse {
+    pub message: String,
+}
+
+// ── Doc cache ────────────────────────────────────────────────────────────────
+
+static DOC_CONTEXT: OnceLock<String> = OnceLock::new();
+
+/// Walk upward from `start` until we find a directory containing both
+/// `README.md` and a `docs/` subdirectory.
+fn find_project_root(start: &Path) -> Option<PathBuf> {
+    let mut current = start.to_path_buf();
+    loop {
+        if current.join("README.md").is_file() && current.join("docs").is_dir() {
+            return Some(current);
+        }
+        if !current.pop() {
+            return None;
+        }
+    }
+}
+
+/// Read README.md + all docs/**/*.md (excluding node_modules).
+fn load_docs(root: &Path) -> String {
+    let mut parts: Vec<String> = Vec::new();
+
+    // Root README first
+    if let Ok(content) = std::fs::read_to_string(root.join("README.md")) {
+        parts.push(format!("<!-- file: README.md -->\n{content}"));
+    }
+
+    // docs/**/*.md, skipping node_modules
+    for entry in WalkDir::new(root.join("docs"))
+        .follow_links(false)
+        .into_iter()
+        .filter_entry(|e| {
+            !e.path()
+                .components()
+                .any(|c| c.as_os_str() == "node_modules")
+        })
+        .filter_map(|e| e.ok())
+    {
+        let path = entry.path();
+        if !path.is_file() {
+            continue;
+        }
+        if path
+            .extension()
+            .and_then(|s| s.to_str())
+            .map(|s| !s.eq_ignore_ascii_case("md"))
+            .unwrap_or(true)
+        {
+            continue;
+        }
+
+        let rel = path.strip_prefix(root).unwrap_or(path);
+        if let Ok(content) = std::fs::read_to_string(path) {
+            parts.push(format!("<!-- file: {} -->\n{content}", rel.display()));
+        }
+    }
+
+    if parts.is_empty() {
+        tracing::warn!(
+            "help_chat: no documentation files found under {}",
+            root.display()
+        );
+    } else {
+        tracing::info!(
+            "help_chat: loaded {} documentation file(s) from {}",
+            parts.len(),
+            root.display()
+        );
+    }
+
+    parts.join("\n\n---\n\n")
+}
+
+/// Returns a reference to the cached doc context string, initialised on
+/// first call via `OnceLock`.
+///
+/// Discovery order:
+/// 1. `HELP_DOCS_PATH` env var (explicit override)
+/// 2. Walk up from the binary location
+/// 3. Current working directory
+/// 4. Common Docker paths (/app, /opt/compliance-scanner)
+fn doc_context() -> &'static str {
+    DOC_CONTEXT.get_or_init(|| {
+        // 1. Explicit env var
+        if let Ok(path) = std::env::var("HELP_DOCS_PATH") {
+            let p = PathBuf::from(&path);
+            if p.join("README.md").is_file() || p.join("docs").is_dir() {
+                tracing::info!("help_chat: loading docs from HELP_DOCS_PATH={path}");
+                return load_docs(&p);
+            }
+            tracing::warn!("help_chat: HELP_DOCS_PATH={path} has no README.md or docs/");
+        }
+
+        // 2. Walk up from binary location
+        let start = std::env::current_exe()
+            .ok()
+            .and_then(|p| p.parent().map(Path::to_path_buf))
+            .unwrap_or_else(|| PathBuf::from("."));
+
+        if let Some(root) = find_project_root(&start) {
+            return load_docs(&root);
+        }
+
+        // 3. Current working directory
+        if let Ok(cwd) = std::env::current_dir() {
+            if let Some(root) = find_project_root(&cwd) {
+                return load_docs(&root);
+            }
+            if cwd.join("README.md").is_file() {
+                return load_docs(&cwd);
+            }
+        }
+
+        // 4. Common Docker/deployment paths
+        for candidate in ["/app", "/opt/compliance-scanner", "/srv/compliance-scanner"] {
+            let p = PathBuf::from(candidate);
+            if p.join("README.md").is_file() || p.join("docs").is_dir() {
+                tracing::info!("help_chat: found docs at {candidate}");
+                return load_docs(&p);
+            }
+        }
+
+        tracing::error!(
+            "help_chat: could not locate project root; doc context will be empty. \
+             Set HELP_DOCS_PATH to the directory containing README.md and docs/"
+        );
+        String::new()
+    })
+}
+
+// ── Handler ──────────────────────────────────────────────────────────────────
+
+/// POST /api/v1/help/chat — Answer questions about the compliance-scanner
+/// using the project documentation as grounding context.
+#[tracing::instrument(skip_all)]
+pub async fn help_chat(
+    Extension(agent): AgentExt,
+    Json(req): Json<HelpChatRequest>,
+) -> Result<Json<ApiResponse<HelpChatResponse>>, StatusCode> {
+    let context = doc_context();
+
+    let system_prompt = if context.is_empty() {
+        "You are a helpful assistant for the Compliance Scanner project. \
+         Answer questions about how to use and configure it. \
+         No documentation was loaded at startup, so rely on your general knowledge."
+            .to_string()
+    } else {
+        format!(
+            "You are a helpful assistant for the Compliance Scanner project. \
+             Answer questions about how to use, configure, and understand it \
+             using the documentation below as your primary source of truth.\n\n\
+             Rules:\n\
+             - Prefer information from the provided docs over general knowledge\n\
+             - Quote or reference the relevant doc section when it helps\n\
+             - If the docs do not cover the topic, say so clearly\n\
+             - Be concise — lead with the answer, then explain if needed\n\
+             - Use markdown formatting for readability\n\n\
+             ## Project Documentation\n\n{context}"
+        )
+    };
+
+    let mut messages: Vec<(String, String)> = Vec::with_capacity(req.history.len() + 2);
+    messages.push(("system".to_string(), system_prompt));
+
+    for msg in &req.history {
+        messages.push((msg.role.clone(), msg.content.clone()));
+    }
+    messages.push(("user".to_string(), req.message));
+
+    let response_text = agent
+        .llm
+        .chat_with_messages(messages, Some(0.3))
+        .await
+        .map_err(|e| {
+            tracing::error!("LLM help chat failed: {e}");
+            StatusCode::INTERNAL_SERVER_ERROR
+        })?;
+
+    Ok(Json(ApiResponse {
+        data: HelpChatResponse {
+            message: response_text,
+        },
+        total: None,
+        page: None,
+    }))
+}

compliance-agent/src/api/handlers/issues.rs

@@ -4,13 +4,16 @@ use mongodb::bson::doc;
 
 use super::dto::*;
 use compliance_core::models::TrackerIssue;
+use compliance_core::tenant_ctx::TenantCtx;
 
 #[tracing::instrument(skip_all)]
 pub async fn list_issues(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Query(params): Query<PaginationParams>,
 ) -> ApiResult<Vec<TrackerIssue>> {
-    let db = &agent.db;
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
     let skip = (params.page.saturating_sub(1)) * params.limit as u64;
     let total = db
         .tracker_issues()

compliance-agent/src/api/handlers/mcp_tokens.rs

@@ -0,0 +1,186 @@
+//! `/api/v1/mcp-tokens` — per-tenant API tokens for the MCP server.
+//!
+//! These are opaque static bearers issued via the dashboard (or a
+//! direct curl with a KC JWT) and copied into LLM clients (Claude
+//! Desktop / Cursor / ChatGPT). The MCP server hashes incoming bearers
+//! and looks them up in the cross-tenant `<prefix>__admin.mcp_tokens`
+//! collection to derive the tenant_id for routing.
+//!
+//! The raw token is shown to the caller exactly once at creation; the
+//! database only ever stores the SHA-256 hash. Revocation is a soft
+//! delete (sets `revoked: true`) so the audit log keeps the record.
+
+use axum::extract::{Extension, Path};
+use axum::http::StatusCode;
+use axum::Json;
+use base64::{engine::general_purpose::URL_SAFE_NO_PAD, Engine as _};
+use compliance_core::models::{McpToken, McpTokenView};
+use compliance_core::tenant_ctx::TenantCtx;
+use mongodb::bson::doc;
+use rand::RngCore;
+use sha2::{Digest, Sha256};
+
+use super::dto::{AgentExt, ApiResponse};
+
+/// Mongo collection name inside the admin DB.
+const COLLECTION: &str = "mcp_tokens";
+
+/// Token prefix the MCP server expects on every bearer.
+const TOKEN_PREFIX: &str = "mcpt_";
+
+/// Bytes of randomness behind each token. 32 → ~256 bits.
+/// Encoded as URL-safe base64 without padding → 43 chars.
+/// Combined with `mcpt_` → 48-char tokens.
+const TOKEN_RAND_BYTES: usize = 32;
+
+#[derive(serde::Deserialize)]
+pub struct CreateMcpTokenRequest {
+    pub name: String,
+}
+
+/// Returned exactly once at creation. The `token` field is gone from
+/// the listing endpoint — the user must save it now.
+#[derive(serde::Serialize)]
+pub struct CreateMcpTokenResponse {
+    pub token: String,
+    pub view: McpTokenView,
+}
+
+/// `POST /api/v1/mcp-tokens` — mint a new token for the caller's tenant.
+#[tracing::instrument(skip_all)]
+pub async fn create_mcp_token(
+    Extension(agent): AgentExt,
+    tenant: TenantCtx,
+    Json(req): Json<CreateMcpTokenRequest>,
+) -> Result<Json<CreateMcpTokenResponse>, StatusCode> {
+    if req.name.trim().is_empty() {
+        return Err(StatusCode::BAD_REQUEST);
+    }
+    let raw = generate_token();
+    let token_hash = sha256_hex(&raw);
+    let token_prefix: String = raw.chars().take(12).collect();
+
+    let mut token = McpToken {
+        id: None,
+        token_hash,
+        token_prefix,
+        tenant_id: tenant.0.tenant_id.clone(),
+        name: req.name.trim().to_string(),
+        created_by: tenant.0.user_id.clone(),
+        created_at: chrono::Utc::now(),
+        last_used_at: None,
+        revoked: false,
+    };
+
+    let col = agent.db_pool.admin_db().collection::<McpToken>(COLLECTION);
+    let res = col.insert_one(&token).await.map_err(|e| {
+        tracing::error!("Failed to insert MCP token: {e}");
+        StatusCode::INTERNAL_SERVER_ERROR
+    })?;
+    token.id = res.inserted_id.as_object_id();
+
+    Ok(Json(CreateMcpTokenResponse {
+        view: McpTokenView::from(&token),
+        token: raw,
+    }))
+}
+
+/// `GET /api/v1/mcp-tokens` — list tokens for the caller's tenant.
+/// Hash is never returned; only metadata + the 12-char prefix so the
+/// user can identify which row is which.
+#[tracing::instrument(skip_all)]
+pub async fn list_mcp_tokens(
+    Extension(agent): AgentExt,
+    tenant: TenantCtx,
+) -> Result<Json<ApiResponse<Vec<McpTokenView>>>, StatusCode> {
+    let col = agent.db_pool.admin_db().collection::<McpToken>(COLLECTION);
+    let mut cursor = col
+        .find(doc! { "tenant_id": &tenant.0.tenant_id })
+        .sort(doc! { "created_at": -1 })
+        .await
+        .map_err(|e| {
+            tracing::error!("Failed to list MCP tokens: {e}");
+            StatusCode::INTERNAL_SERVER_ERROR
+        })?;
+    let mut out = Vec::new();
+    while cursor.advance().await.map_err(|e| {
+        tracing::warn!("MCP tokens cursor advance failed: {e}");
+        StatusCode::INTERNAL_SERVER_ERROR
+    })? {
+        match cursor.deserialize_current() {
+            Ok(t) => out.push(McpTokenView::from(&t)),
+            Err(e) => tracing::warn!("Failed to deserialize MCP token: {e}"),
+        }
+    }
+    Ok(Json(ApiResponse {
+        data: out,
+        total: None,
+        page: None,
+    }))
+}
+
+/// `DELETE /api/v1/mcp-tokens/{id}` — revoke (soft delete).
+/// Scoped to the caller's tenant: a user can't revoke another tenant's
+/// token even if they guess its id.
+#[tracing::instrument(skip_all, fields(id = %id))]
+pub async fn revoke_mcp_token(
+    Extension(agent): AgentExt,
+    tenant: TenantCtx,
+    Path(id): Path<String>,
+) -> Result<Json<serde_json::Value>, StatusCode> {
+    let oid = mongodb::bson::oid::ObjectId::parse_str(&id).map_err(|_| StatusCode::BAD_REQUEST)?;
+    let col = agent.db_pool.admin_db().collection::<McpToken>(COLLECTION);
+    let result = col
+        .update_one(
+            doc! { "_id": oid, "tenant_id": &tenant.0.tenant_id },
+            doc! { "$set": { "revoked": true } },
+        )
+        .await
+        .map_err(|e| {
+            tracing::error!("Failed to revoke MCP token: {e}");
+            StatusCode::INTERNAL_SERVER_ERROR
+        })?;
+    if result.matched_count == 0 {
+        return Err(StatusCode::NOT_FOUND);
+    }
+    Ok(Json(serde_json::json!({ "status": "revoked" })))
+}
+
+/// 32 bytes random → URL-safe base64 → 43 chars, no padding.
+/// Prefixed with `mcpt_` so the MCP server can sniff the format
+/// before bothering with the DB lookup.
+fn generate_token() -> String {
+    let mut bytes = [0u8; TOKEN_RAND_BYTES];
+    rand::rng().fill_bytes(&mut bytes);
+    format!("{TOKEN_PREFIX}{}", URL_SAFE_NO_PAD.encode(bytes))
+}
+
+fn sha256_hex(s: &str) -> String {
+    let mut h = Sha256::new();
+    h.update(s.as_bytes());
+    hex::encode(h.finalize())
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn generated_tokens_are_unique_and_prefixed() {
+        let a = generate_token();
+        let b = generate_token();
+        assert_ne!(a, b);
+        assert!(a.starts_with(TOKEN_PREFIX));
+        assert!(b.starts_with(TOKEN_PREFIX));
+        // 5 + 43 = 48 chars
+        assert_eq!(a.len(), 5 + 43);
+    }
+
+    #[test]
+    fn sha256_is_stable_and_64_hex() {
+        let h = sha256_hex("mcpt_abc");
+        assert_eq!(h.len(), 64);
+        assert!(h.chars().all(|c| c.is_ascii_hexdigit()));
+        assert_eq!(sha256_hex("mcpt_abc"), h);
+    }
+}

compliance-agent/src/api/handlers/mod.rs

@@ -4,7 +4,10 @@ pub mod dto;
 pub mod findings;
 pub mod graph;
 pub mod health;
+pub mod help_chat;
 pub mod issues;
+pub mod mcp_tokens;
+pub mod notifications;
 pub mod pentest_handlers;
 pub use pentest_handlers as pentest;
 pub mod repos;

compliance-agent/src/api/handlers/notifications.rs

@@ -0,0 +1,183 @@
+use axum::extract::Extension;
+use axum::http::StatusCode;
+use axum::Json;
+use mongodb::bson::doc;
+use serde::Deserialize;
+
+use compliance_core::models::notification::CveNotification;
+use compliance_core::tenant_ctx::TenantCtx;
+
+use super::dto::{tenant_db, AgentExt, ApiResponse};
+
+/// GET /api/v1/notifications — List CVE notifications (newest first)
+#[tracing::instrument(skip_all)]
+pub async fn list_notifications(
+    Extension(agent): AgentExt,
+    tenant: TenantCtx,
+    axum::extract::Query(params): axum::extract::Query<NotificationFilter>,
+) -> Result<Json<ApiResponse<Vec<CveNotification>>>, StatusCode> {
+    let db = tenant_db(&agent, &tenant).await?;
+    let mut filter = doc! {};
+
+    // Filter by status (default: show new + read, exclude dismissed)
+    match params.status.as_deref() {
+        Some("all") => {}
+        Some(s) => {
+            filter.insert("status", s);
+        }
+        None => {
+            filter.insert("status", doc! { "$in": ["new", "read"] });
+        }
+    }
+
+    // Filter by severity
+    if let Some(ref sev) = params.severity {
+        filter.insert("severity", sev.as_str());
+    }
+
+    // Filter by repo
+    if let Some(ref repo_id) = params.repo_id {
+        filter.insert("repo_id", repo_id.as_str());
+    }
+
+    let page = params.page.unwrap_or(1).max(1);
+    let limit = params.limit.unwrap_or(50).min(200);
+    let skip = (page - 1) * limit as u64;
+
+    let total = db
+        .cve_notifications()
+        .count_documents(filter.clone())
+        .await
+        .unwrap_or(0);
+
+    let notifications: Vec<CveNotification> = match db
+        .cve_notifications()
+        .find(filter)
+        .sort(doc! { "created_at": -1 })
+        .skip(skip)
+        .limit(limit)
+        .await
+    {
+        Ok(cursor) => {
+            use futures_util::StreamExt;
+            let mut items = Vec::new();
+            let mut cursor = cursor;
+            while let Some(Ok(n)) = cursor.next().await {
+                items.push(n);
+            }
+            items
+        }
+        Err(e) => {
+            tracing::error!("Failed to list notifications: {e}");
+            return Err(StatusCode::INTERNAL_SERVER_ERROR);
+        }
+    };
+
+    Ok(Json(ApiResponse {
+        data: notifications,
+        total: Some(total),
+        page: Some(page),
+    }))
+}
+
+/// GET /api/v1/notifications/count — Count of unread notifications
+#[tracing::instrument(skip_all)]
+pub async fn notification_count(
+    Extension(agent): AgentExt,
+    tenant: TenantCtx,
+) -> Result<Json<serde_json::Value>, StatusCode> {
+    let db = tenant_db(&agent, &tenant).await?;
+    let count = db
+        .cve_notifications()
+        .count_documents(doc! { "status": "new" })
+        .await
+        .unwrap_or(0);
+
+    Ok(Json(serde_json::json!({ "count": count })))
+}
+
+/// PATCH /api/v1/notifications/:id/read — Mark a notification as read
+#[tracing::instrument(skip_all, fields(id = %id))]
+pub async fn mark_read(
+    Extension(agent): AgentExt,
+    tenant: TenantCtx,
+    axum::extract::Path(id): axum::extract::Path<String>,
+) -> Result<Json<serde_json::Value>, StatusCode> {
+    let oid = mongodb::bson::oid::ObjectId::parse_str(&id).map_err(|_| StatusCode::BAD_REQUEST)?;
+    let db = tenant_db(&agent, &tenant).await?;
+
+    let result = db
+        .cve_notifications()
+        .update_one(
+            doc! { "_id": oid },
+            doc! { "$set": {
+                "status": "read",
+                "read_at": mongodb::bson::DateTime::now(),
+            }},
+        )
+        .await
+        .map_err(|_| StatusCode::INTERNAL_SERVER_ERROR)?;
+
+    if result.matched_count == 0 {
+        return Err(StatusCode::NOT_FOUND);
+    }
+    Ok(Json(serde_json::json!({ "status": "read" })))
+}
+
+/// PATCH /api/v1/notifications/:id/dismiss — Dismiss a notification
+#[tracing::instrument(skip_all, fields(id = %id))]
+pub async fn dismiss_notification(
+    Extension(agent): AgentExt,
+    tenant: TenantCtx,
+    axum::extract::Path(id): axum::extract::Path<String>,
+) -> Result<Json<serde_json::Value>, StatusCode> {
+    let oid = mongodb::bson::oid::ObjectId::parse_str(&id).map_err(|_| StatusCode::BAD_REQUEST)?;
+    let db = tenant_db(&agent, &tenant).await?;
+
+    let result = db
+        .cve_notifications()
+        .update_one(
+            doc! { "_id": oid },
+            doc! { "$set": { "status": "dismissed" } },
+        )
+        .await
+        .map_err(|_| StatusCode::INTERNAL_SERVER_ERROR)?;
+
+    if result.matched_count == 0 {
+        return Err(StatusCode::NOT_FOUND);
+    }
+    Ok(Json(serde_json::json!({ "status": "dismissed" })))
+}
+
+/// POST /api/v1/notifications/read-all — Mark all new notifications as read
+#[tracing::instrument(skip_all)]
+pub async fn mark_all_read(
+    Extension(agent): AgentExt,
+    tenant: TenantCtx,
+) -> Result<Json<serde_json::Value>, StatusCode> {
+    let db = tenant_db(&agent, &tenant).await?;
+    let result = db
+        .cve_notifications()
+        .update_many(
+            doc! { "status": "new" },
+            doc! { "$set": {
+                "status": "read",
+                "read_at": mongodb::bson::DateTime::now(),
+            }},
+        )
+        .await
+        .map_err(|_| StatusCode::INTERNAL_SERVER_ERROR)?;
+
+    Ok(Json(
+        serde_json::json!({ "updated": result.modified_count }),
+    ))
+}
+
+#[derive(Debug, Deserialize)]
+pub struct NotificationFilter {
+    pub status: Option<String>,
+    pub severity: Option<String>,
+    pub repo_id: Option<String>,
+    pub page: Option<u64>,
+    pub limit: Option<i64>,
+}

compliance-agent/src/api/handlers/pentest_handlers/export.rs

@@ -13,10 +13,11 @@ use compliance_core::models::dast::DastFinding;
 use compliance_core::models::finding::Finding;
 use compliance_core::models::pentest::*;
 use compliance_core::models::sbom::SbomEntry;
+use compliance_core::tenant_ctx::TenantCtx;
 
 use crate::agent::ComplianceAgent;
 
-use super::super::dto::collect_cursor_async;
+use super::super::dto::{collect_cursor_async, tenant_db};
 
 type AgentExt = Extension<Arc<ComplianceAgent>>;
 
@@ -35,11 +36,15 @@ pub struct ExportBody {
 #[tracing::instrument(skip_all, fields(session_id = %id))]
 pub async fn export_session_report(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(id): Path<String>,
     Json(body): Json<ExportBody>,
 ) -> Result<axum::response::Response, (StatusCode, String)> {
     let oid = mongodb::bson::oid::ObjectId::parse_str(&id)
         .map_err(|_| (StatusCode::BAD_REQUEST, "Invalid session ID".to_string()))?;
+    let db = tenant_db(&agent, &tenant)
+        .await
+        .map_err(|s| (s, "failed to acquire tenant database".to_string()))?;
 
     if body.password.len() < 8 {
         return Err((
@@ -49,8 +54,7 @@ pub async fn export_session_report(
     }
 
     // Fetch session
-    let session = agent
-        .db
+    let session = db
         .pentest_sessions()
         .find_one(doc! { "_id": oid })
         .await
@@ -64,9 +68,7 @@ pub async fn export_session_report(
 
     // Resolve target name
     let target = if let Ok(tid) = mongodb::bson::oid::ObjectId::parse_str(&session.target_id) {
-        agent
-            .db
-            .dast_targets()
+        db.dast_targets()
             .find_one(doc! { "_id": tid })
             .await
             .ok()
@@ -84,8 +86,7 @@ pub async fn export_session_report(
         .unwrap_or_default();
 
     // Fetch attack chain nodes
-    let nodes: Vec<AttackChainNode> = match agent
-        .db
+    let nodes: Vec<AttackChainNode> = match db
         .attack_chain_nodes()
         .find(doc! { "session_id": &id })
         .sort(doc! { "started_at": 1 })
@@ -95,9 +96,8 @@ pub async fn export_session_report(
         Err(_) => Vec::new(),
     };
 
-    // Fetch DAST findings for this session
-    let findings: Vec<DastFinding> = match agent
-        .db
+    // Fetch DAST findings for this session, then deduplicate
+    let raw_findings: Vec<DastFinding> = match db
         .dast_findings()
         .find(doc! { "session_id": &id })
         .sort(doc! { "severity": -1, "created_at": -1 })
@@ -106,6 +106,14 @@ pub async fn export_session_report(
         Ok(cursor) => collect_cursor_async(cursor).await,
         Err(_) => Vec::new(),
     };
+    let raw_count = raw_findings.len();
+    let findings = crate::pipeline::dedup::dedup_dast_findings(raw_findings);
+    if findings.len() < raw_count {
+        tracing::info!(
+            "Deduped DAST findings for session {id}: {raw_count} → {}",
+            findings.len()
+        );
+    }
 
     // Fetch SAST findings, SBOM, and code context for the linked repository
     let repo_id = session
@@ -114,8 +122,7 @@ pub async fn export_session_report(
         .or_else(|| target.as_ref().and_then(|t| t.repo_id.clone()));
 
     let (sast_findings, sbom_entries, code_context) = if let Some(ref rid) = repo_id {
-        let sast: Vec<Finding> = match agent
-            .db
+        let sast: Vec<Finding> = match db
             .findings()
             .find(doc! {
                 "repo_id": rid,
@@ -135,8 +142,7 @@ pub async fn export_session_report(
             Err(_) => Vec::new(),
         };
 
-        let sbom: Vec<SbomEntry> = match agent
-            .db
+        let sbom: Vec<SbomEntry> = match db
             .sbom_entries()
             .find(doc! {
                 "repo_id": rid,
@@ -156,8 +162,7 @@ pub async fn export_session_report(
         };
 
         // Build code context from graph nodes
-        let code_ctx: Vec<CodeContextHint> = match agent
-            .db
+        let code_ctx: Vec<CodeContextHint> = match db
             .graph_nodes()
             .find(doc! { "repo_id": rid, "is_entry_point": true })
             .limit(50)

compliance-agent/src/api/handlers/pentest_handlers/session.rs

@@ -7,11 +7,12 @@ use mongodb::bson::doc;
 use serde::Deserialize;
 
 use compliance_core::models::pentest::*;
+use compliance_core::tenant_ctx::TenantCtx;
 
 use crate::agent::ComplianceAgent;
 use crate::pentest::PentestOrchestrator;
 
-use super::super::dto::{collect_cursor_async, ApiResponse, PaginationParams};
+use super::super::dto::{collect_cursor_async, tenant_db, ApiResponse, PaginationParams};
 
 type AgentExt = Extension<Arc<ComplianceAgent>>;
 
@@ -43,6 +44,7 @@ pub struct LookupRepoQuery {
 #[tracing::instrument(skip_all)]
 pub async fn create_session(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Json(req): Json<CreateSessionRequest>,
 ) -> Result<Json<ApiResponse<PentestSession>>, (StatusCode, String)> {
     // Try to acquire a concurrency permit
@@ -57,6 +59,10 @@ pub async fn create_session(
             )
         })?;
 
+    let db = tenant_db(&agent, &tenant)
+        .await
+        .map_err(|s| (s, "failed to acquire tenant database".to_string()))?;
+
     if let Some(ref config) = req.config {
         // ── Wizard path ──────────────────────────────────────────────
         if !config.disclaimer_accepted {
@@ -67,8 +73,7 @@ pub async fn create_session(
         }
 
         // Look up or auto-create DastTarget by app_url
-        let target = match agent
-            .db
+        let target = match db
             .dast_targets()
             .find_one(doc! { "base_url": &config.app_url })
             .await
@@ -87,7 +92,7 @@ pub async fn create_session(
                 }
                 t.allow_destructive = config.allow_destructive;
                 t.excluded_paths = config.scope_exclusions.clone();
-                let res = agent.db.dast_targets().insert_one(&t).await.map_err(|e| {
+                let res = db.dast_targets().insert_one(&t).await.map_err(|e| {
                     (
                         StatusCode::INTERNAL_SERVER_ERROR,
                         format!("Failed to create target: {e}"),
@@ -110,8 +115,7 @@ pub async fn create_session(
 
         // Resolve repo_id from git_repo_url if provided
         if let Some(ref git_url) = config.git_repo_url {
-            if let Ok(Some(repo)) = agent
-                .db
+            if let Ok(Some(repo)) = db
                 .repositories()
                 .find_one(doc! { "git_url": git_url })
                 .await
@@ -120,8 +124,7 @@ pub async fn create_session(
             }
         }
 
-        let insert_result = agent
-            .db
+        let insert_result = db
             .pentest_sessions()
             .insert_one(&session)
             .await
@@ -212,8 +215,7 @@ pub async fn create_session(
         // Persist encrypted credentials to DB
         if session_for_task.config.is_some() {
             if let Some(sid) = session.id {
-                let _ = agent
-                    .db
+                let _ = db
                     .pentest_sessions()
                     .update_one(
                         doc! { "_id": sid },
@@ -245,12 +247,13 @@ pub async fn create_session(
             });
 
         let llm = agent.llm.clone();
-        let db = agent.db.clone();
+        let db_for_orchestrator = db.clone();
         let session_clone = session.clone();
         let target_clone = target.clone();
         let agent_ref = agent.clone();
         tokio::spawn(async move {
-            let orchestrator = PentestOrchestrator::new(llm, db, event_tx, Some(pause_rx));
+            let orchestrator =
+                PentestOrchestrator::new(llm, db_for_orchestrator, event_tx, Some(pause_rx));
             orchestrator
                 .run_session_guarded(&session_clone, &target_clone, &initial_message)
                 .await;
@@ -292,8 +295,7 @@ pub async fn create_session(
             )
         })?;
 
-        let target = agent
-            .db
+        let target = db
             .dast_targets()
             .find_one(doc! { "_id": oid })
             .await
@@ -310,8 +312,7 @@ pub async fn create_session(
         let mut session = PentestSession::new(target_id, strategy);
         session.repo_id = target.repo_id.clone();
 
-        let insert_result = agent
-            .db
+        let insert_result = db
             .pentest_sessions()
             .insert_one(&session)
             .await
@@ -338,12 +339,13 @@ pub async fn create_session(
         });
 
         let llm = agent.llm.clone();
-        let db = agent.db.clone();
+        let db_for_orchestrator = db.clone();
         let session_clone = session.clone();
         let target_clone = target.clone();
         let agent_ref = agent.clone();
         tokio::spawn(async move {
-            let orchestrator = PentestOrchestrator::new(llm, db, event_tx, Some(pause_rx));
+            let orchestrator =
+                PentestOrchestrator::new(llm, db_for_orchestrator, event_tx, Some(pause_rx));
             orchestrator
                 .run_session_guarded(&session_clone, &target_clone, &initial_message)
                 .await;
@@ -373,10 +375,11 @@ fn parse_strategy(s: &str) -> PentestStrategy {
 #[tracing::instrument(skip_all)]
 pub async fn lookup_repo(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Query(params): Query<LookupRepoQuery>,
 ) -> Result<Json<ApiResponse<serde_json::Value>>, StatusCode> {
-    let repo = agent
-        .db
+    let db = tenant_db(&agent, &tenant).await?;
+    let repo = db
         .repositories()
         .find_one(doc! { "git_url": &params.url })
         .await
@@ -402,9 +405,11 @@ pub async fn lookup_repo(
 #[tracing::instrument(skip_all)]
 pub async fn list_sessions(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Query(params): Query<PaginationParams>,
 ) -> Result<Json<ApiResponse<Vec<PentestSession>>>, StatusCode> {
-    let db = &agent.db;
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
     let skip = (params.page.saturating_sub(1)) * params.limit as u64;
     let total = db
         .pentest_sessions()
@@ -438,12 +443,13 @@ pub async fn list_sessions(
 #[tracing::instrument(skip_all, fields(session_id = %id))]
 pub async fn get_session(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(id): Path<String>,
 ) -> Result<Json<ApiResponse<PentestSession>>, StatusCode> {
     let oid = mongodb::bson::oid::ObjectId::parse_str(&id).map_err(|_| StatusCode::BAD_REQUEST)?;
+    let db = tenant_db(&agent, &tenant).await?;
 
-    let mut session = agent
-        .db
+    let mut session = db
         .pentest_sessions()
         .find_one(doc! { "_id": oid })
         .await
@@ -471,15 +477,18 @@ pub async fn get_session(
 #[tracing::instrument(skip_all, fields(session_id = %id))]
 pub async fn send_message(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(id): Path<String>,
     Json(req): Json<SendMessageRequest>,
 ) -> Result<Json<ApiResponse<PentestMessage>>, (StatusCode, String)> {
     let oid = mongodb::bson::oid::ObjectId::parse_str(&id)
         .map_err(|_| (StatusCode::BAD_REQUEST, "Invalid session ID".to_string()))?;
+    let db = tenant_db(&agent, &tenant)
+        .await
+        .map_err(|s| (s, "failed to acquire tenant database".to_string()))?;
 
     // Verify session exists and is running
-    let session = agent
-        .db
+    let session = db
         .pentest_sessions()
         .find_one(doc! { "_id": oid })
         .await
@@ -506,8 +515,7 @@ pub async fn send_message(
         )
     })?;
 
-    let target = agent
-        .db
+    let target = db
         .dast_targets()
         .find_one(doc! { "_id": target_oid })
         .await
@@ -527,13 +535,13 @@ pub async fn send_message(
     // Store user message
     let session_id = id.clone();
     let user_msg = PentestMessage::user(session_id.clone(), req.message.clone());
-    let _ = agent.db.pentest_messages().insert_one(&user_msg).await;
+    let _ = db.pentest_messages().insert_one(&user_msg).await;
 
     let response_msg = user_msg.clone();
 
     // Spawn orchestrator to continue the session
     let llm = agent.llm.clone();
-    let db = agent.db.clone();
+    let db_for_orchestrator = db.clone();
     let message = req.message.clone();
 
     // Use existing broadcast sender if available, otherwise create a new one
@@ -548,7 +556,7 @@ pub async fn send_message(
         .unwrap_or_else(|| agent.register_session_stream(&session_id));
 
     tokio::spawn(async move {
-        let orchestrator = PentestOrchestrator::new(llm, db, event_tx, None);
+        let orchestrator = PentestOrchestrator::new(llm, db_for_orchestrator, event_tx, None);
         orchestrator
             .run_session_guarded(&session, &target, &message)
             .await;
@@ -565,13 +573,16 @@ pub async fn send_message(
 #[tracing::instrument(skip_all, fields(session_id = %id))]
 pub async fn stop_session(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(id): Path<String>,
 ) -> Result<Json<ApiResponse<PentestSession>>, (StatusCode, String)> {
     let oid = mongodb::bson::oid::ObjectId::parse_str(&id)
         .map_err(|_| (StatusCode::BAD_REQUEST, "Invalid session ID".to_string()))?;
+    let db = tenant_db(&agent, &tenant)
+        .await
+        .map_err(|s| (s, "failed to acquire tenant database".to_string()))?;
 
-    let session = agent
-        .db
+    let session = db
         .pentest_sessions()
         .find_one(doc! { "_id": oid })
         .await
@@ -590,9 +601,7 @@ pub async fn stop_session(
         ));
     }
 
-    agent
-        .db
-        .pentest_sessions()
+    db.pentest_sessions()
         .update_one(
             doc! { "_id": oid },
             doc! { "$set": {
@@ -612,8 +621,7 @@ pub async fn stop_session(
     // Clean up session resources
     agent.cleanup_session(&id);
 
-    let updated = agent
-        .db
+    let updated = db
         .pentest_sessions()
         .find_one(doc! { "_id": oid })
         .await
@@ -641,13 +649,16 @@ pub async fn stop_session(
 #[tracing::instrument(skip_all, fields(session_id = %id))]
 pub async fn pause_session(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(id): Path<String>,
 ) -> Result<Json<ApiResponse<serde_json::Value>>, (StatusCode, String)> {
     let oid = mongodb::bson::oid::ObjectId::parse_str(&id)
         .map_err(|_| (StatusCode::BAD_REQUEST, "Invalid session ID".to_string()))?;
+    let db = tenant_db(&agent, &tenant)
+        .await
+        .map_err(|s| (s, "failed to acquire tenant database".to_string()))?;
 
-    let session = agent
-        .db
+    let session = db
         .pentest_sessions()
         .find_one(doc! { "_id": oid })
         .await
@@ -684,13 +695,16 @@ pub async fn pause_session(
 #[tracing::instrument(skip_all, fields(session_id = %id))]
 pub async fn resume_session(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(id): Path<String>,
 ) -> Result<Json<ApiResponse<serde_json::Value>>, (StatusCode, String)> {
     let oid = mongodb::bson::oid::ObjectId::parse_str(&id)
         .map_err(|_| (StatusCode::BAD_REQUEST, "Invalid session ID".to_string()))?;
+    let db = tenant_db(&agent, &tenant)
+        .await
+        .map_err(|s| (s, "failed to acquire tenant database".to_string()))?;
 
-    let session = agent
-        .db
+    let session = db
         .pentest_sessions()
         .find_one(doc! { "_id": oid })
         .await
@@ -727,12 +741,13 @@ pub async fn resume_session(
 #[tracing::instrument(skip_all, fields(session_id = %id))]
 pub async fn get_attack_chain(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(id): Path<String>,
 ) -> Result<Json<ApiResponse<Vec<AttackChainNode>>>, StatusCode> {
     let _oid = mongodb::bson::oid::ObjectId::parse_str(&id).map_err(|_| StatusCode::BAD_REQUEST)?;
+    let db = tenant_db(&agent, &tenant).await?;
 
-    let nodes = match agent
-        .db
+    let nodes = match db
         .attack_chain_nodes()
         .find(doc! { "session_id": &id })
         .sort(doc! { "started_at": 1 })
@@ -757,21 +772,21 @@ pub async fn get_attack_chain(
 #[tracing::instrument(skip_all, fields(session_id = %id))]
 pub async fn get_messages(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(id): Path<String>,
     Query(params): Query<PaginationParams>,
 ) -> Result<Json<ApiResponse<Vec<PentestMessage>>>, StatusCode> {
     let _oid = mongodb::bson::oid::ObjectId::parse_str(&id).map_err(|_| StatusCode::BAD_REQUEST)?;
+    let db = tenant_db(&agent, &tenant).await?;
 
     let skip = (params.page.saturating_sub(1)) * params.limit as u64;
-    let total = agent
-        .db
+    let total = db
         .pentest_messages()
         .count_documents(doc! { "session_id": &id })
         .await
         .unwrap_or(0);
 
-    let messages = match agent
-        .db
+    let messages = match db
         .pentest_messages()
         .find(doc! { "session_id": &id })
         .sort(doc! { "created_at": 1 })
@@ -797,21 +812,21 @@ pub async fn get_messages(
 #[tracing::instrument(skip_all, fields(session_id = %id))]
 pub async fn get_session_findings(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(id): Path<String>,
     Query(params): Query<PaginationParams>,
 ) -> Result<Json<ApiResponse<Vec<compliance_core::models::dast::DastFinding>>>, StatusCode> {
     let _oid = mongodb::bson::oid::ObjectId::parse_str(&id).map_err(|_| StatusCode::BAD_REQUEST)?;
+    let db = tenant_db(&agent, &tenant).await?;
 
     let skip = (params.page.saturating_sub(1)) * params.limit as u64;
-    let total = agent
-        .db
+    let total = db
         .dast_findings()
         .count_documents(doc! { "session_id": &id })
         .await
         .unwrap_or(0);
 
-    let findings = match agent
-        .db
+    let findings = match db
         .dast_findings()
         .find(doc! { "session_id": &id })
         .sort(doc! { "created_at": -1 })

compliance-agent/src/api/handlers/pentest_handlers/stats.rs

@@ -6,10 +6,11 @@ use axum::Json;
 use mongodb::bson::doc;
 
 use compliance_core::models::pentest::*;
+use compliance_core::tenant_ctx::TenantCtx;
 
 use crate::agent::ComplianceAgent;
 
-use super::super::dto::{collect_cursor_async, ApiResponse};
+use super::super::dto::{collect_cursor_async, tenant_db, ApiResponse};
 
 type AgentExt = Extension<Arc<ComplianceAgent>>;
 
@@ -17,8 +18,10 @@ type AgentExt = Extension<Arc<ComplianceAgent>>;
 #[tracing::instrument(skip_all)]
 pub async fn pentest_stats(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
 ) -> Result<Json<ApiResponse<PentestStats>>, StatusCode> {
-    let db = &agent.db;
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
 
     let running_sessions = db
         .pentest_sessions()

compliance-agent/src/api/handlers/pentest_handlers/stream.rs

@@ -11,10 +11,11 @@ use tokio_stream::wrappers::BroadcastStream;
 use tokio_stream::StreamExt;
 
 use compliance_core::models::pentest::*;
+use compliance_core::tenant_ctx::TenantCtx;
 
 use crate::agent::ComplianceAgent;
 
-use super::super::dto::collect_cursor_async;
+use super::super::dto::{collect_cursor_async, tenant_db};
 
 type AgentExt = Extension<Arc<ComplianceAgent>>;
 
@@ -25,13 +26,14 @@ type AgentExt = Extension<Arc<ComplianceAgent>>;
 #[tracing::instrument(skip_all, fields(session_id = %id))]
 pub async fn session_stream(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(id): Path<String>,
 ) -> Result<Sse<impl futures_util::Stream<Item = Result<Event, Infallible>>>, StatusCode> {
     let oid = mongodb::bson::oid::ObjectId::parse_str(&id).map_err(|_| StatusCode::BAD_REQUEST)?;
+    let db = tenant_db(&agent, &tenant).await?;
 
     // Verify session exists
-    let _session = agent
-        .db
+    let _session = db
         .pentest_sessions()
         .find_one(doc! { "_id": oid })
         .await
@@ -43,8 +45,7 @@ pub async fn session_stream(
     let mut initial_events: Vec<Result<Event, Infallible>> = Vec::new();
 
     // Fetch recent messages for this session
-    let messages: Vec<PentestMessage> = match agent
-        .db
+    let messages: Vec<PentestMessage> = match db
         .pentest_messages()
         .find(doc! { "session_id": &id })
         .sort(doc! { "created_at": 1 })
@@ -56,8 +57,7 @@ pub async fn session_stream(
     };
 
     // Fetch recent attack chain nodes
-    let nodes: Vec<AttackChainNode> = match agent
-        .db
+    let nodes: Vec<AttackChainNode> = match db
         .attack_chain_nodes()
         .find(doc! { "session_id": &id })
         .sort(doc! { "started_at": 1 })
@@ -94,8 +94,7 @@ pub async fn session_stream(
     }
 
     // Add current session status event
-    let session = agent
-        .db
+    let session = db
         .pentest_sessions()
         .find_one(doc! { "_id": oid })
         .await

compliance-agent/src/api/handlers/repos.rs

@@ -5,13 +5,16 @@ use mongodb::bson::doc;
 
 use super::dto::*;
 use compliance_core::models::*;
+use compliance_core::tenant_ctx::TenantCtx;
 
 #[tracing::instrument(skip_all)]
 pub async fn list_repositories(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Query(params): Query<PaginationParams>,
 ) -> ApiResult<Vec<TrackedRepository>> {
-    let db = &agent.db;
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
     let skip = (params.page.saturating_sub(1)) * params.limit as u64;
     let total = db
         .repositories()
@@ -43,6 +46,7 @@ pub async fn list_repositories(
 #[tracing::instrument(skip_all)]
 pub async fn add_repository(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Json(req): Json<AddRepositoryRequest>,
 ) -> Result<Json<ApiResponse<TrackedRepository>>, (StatusCode, String)> {
     // Validate repository access before saving
@@ -69,17 +73,15 @@ pub async fn add_repository(
     repo.tracker_token = req.tracker_token;
     repo.scan_schedule = req.scan_schedule;
 
-    agent
-        .db
-        .repositories()
-        .insert_one(&repo)
+    let db = tenant_db(&agent, &tenant)
         .await
-        .map_err(|_| {
-            (
-                StatusCode::CONFLICT,
-                "Repository already exists".to_string(),
-            )
-        })?;
+        .map_err(|s| (s, "failed to acquire tenant database".to_string()))?;
+    db.repositories().insert_one(&repo).await.map_err(|_| {
+        (
+            StatusCode::CONFLICT,
+            "Repository already exists".to_string(),
+        )
+    })?;
 
     Ok(Json(ApiResponse {
         data: repo,
@@ -91,10 +93,12 @@ pub async fn add_repository(
 #[tracing::instrument(skip_all, fields(repo_id = %id))]
 pub async fn update_repository(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(id): Path<String>,
     Json(req): Json<UpdateRepositoryRequest>,
 ) -> Result<Json<serde_json::Value>, StatusCode> {
     let oid = mongodb::bson::oid::ObjectId::parse_str(&id).map_err(|_| StatusCode::BAD_REQUEST)?;
+    let db = tenant_db(&agent, &tenant).await?;
 
     let mut set_doc = doc! { "updated_at": mongodb::bson::DateTime::now() };
 
@@ -126,8 +130,7 @@ pub async fn update_repository(
         set_doc.insert("scan_schedule", schedule);
     }
 
-    let result = agent
-        .db
+    let result = db
         .repositories()
         .update_one(doc! { "_id": oid }, doc! { "$set": set_doc })
         .await
@@ -155,11 +158,16 @@ pub async fn get_ssh_public_key(
 #[tracing::instrument(skip_all, fields(repo_id = %id))]
 pub async fn trigger_scan(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(id): Path<String>,
 ) -> Result<Json<serde_json::Value>, StatusCode> {
     let agent_clone = (*agent).clone();
+    let tenant_id = tenant.0.tenant_id.clone();
     tokio::spawn(async move {
-        if let Err(e) = agent_clone.run_scan(&id, ScanTrigger::Manual).await {
+        if let Err(e) = agent_clone
+            .run_scan(&tenant_id, &id, ScanTrigger::Manual)
+            .await
+        {
             tracing::error!("Manual scan failed for {id}: {e}");
         }
     });
@@ -170,11 +178,12 @@ pub async fn trigger_scan(
 /// Return the webhook secret for a repository (used by dashboard to display it)
 pub async fn get_webhook_config(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(id): Path<String>,
 ) -> Result<Json<serde_json::Value>, StatusCode> {
     let oid = mongodb::bson::oid::ObjectId::parse_str(&id).map_err(|_| StatusCode::BAD_REQUEST)?;
-    let repo = agent
-        .db
+    let db = tenant_db(&agent, &tenant).await?;
+    let repo = db
         .repositories()
         .find_one(doc! { "_id": oid })
         .await
@@ -196,10 +205,12 @@ pub async fn get_webhook_config(
 #[tracing::instrument(skip_all, fields(repo_id = %id))]
 pub async fn delete_repository(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Path(id): Path<String>,
 ) -> Result<Json<serde_json::Value>, StatusCode> {
     let oid = mongodb::bson::oid::ObjectId::parse_str(&id).map_err(|_| StatusCode::BAD_REQUEST)?;
-    let db = &agent.db;
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
 
     // Delete the repository
     let result = db
@@ -237,5 +248,92 @@ pub async fn delete_repository(
         .delete_many(doc! { "repo_id": &id })
         .await;
 
+    // Cascade delete DAST targets linked to this repo, and all their downstream data
+    // (scan runs, findings, pentest sessions, attack chains, messages)
+    if let Ok(mut cursor) = db.dast_targets().find(doc! { "repo_id": &id }).await {
+        use futures_util::StreamExt;
+        while let Some(Ok(target)) = cursor.next().await {
+            let target_id = target.id.map(|oid| oid.to_hex()).unwrap_or_default();
+            if !target_id.is_empty() {
+                cascade_delete_dast_target(db, &target_id).await;
+            }
+        }
+    }
+
+    // Also delete pentest sessions linked directly to this repo (not via target)
+    if let Ok(mut cursor) = db.pentest_sessions().find(doc! { "repo_id": &id }).await {
+        use futures_util::StreamExt;
+        while let Some(Ok(session)) = cursor.next().await {
+            let session_id = session.id.map(|oid| oid.to_hex()).unwrap_or_default();
+            if !session_id.is_empty() {
+                let _ = db
+                    .attack_chain_nodes()
+                    .delete_many(doc! { "session_id": &session_id })
+                    .await;
+                let _ = db
+                    .pentest_messages()
+                    .delete_many(doc! { "session_id": &session_id })
+                    .await;
+                // Delete DAST findings produced by this session
+                let _ = db
+                    .dast_findings()
+                    .delete_many(doc! { "session_id": &session_id })
+                    .await;
+            }
+        }
+    }
+    let _ = db
+        .pentest_sessions()
+        .delete_many(doc! { "repo_id": &id })
+        .await;
+
     Ok(Json(serde_json::json!({ "status": "deleted" })))
 }
+
+/// Cascade-delete a DAST target and all its downstream data.
+async fn cascade_delete_dast_target(db: &crate::database::Database, target_id: &str) {
+    // Delete pentest sessions for this target (and their attack chains + messages)
+    if let Ok(mut cursor) = db
+        .pentest_sessions()
+        .find(doc! { "target_id": target_id })
+        .await
+    {
+        use futures_util::StreamExt;
+        while let Some(Ok(session)) = cursor.next().await {
+            let session_id = session.id.map(|oid| oid.to_hex()).unwrap_or_default();
+            if !session_id.is_empty() {
+                let _ = db
+                    .attack_chain_nodes()
+                    .delete_many(doc! { "session_id": &session_id })
+                    .await;
+                let _ = db
+                    .pentest_messages()
+                    .delete_many(doc! { "session_id": &session_id })
+                    .await;
+                let _ = db
+                    .dast_findings()
+                    .delete_many(doc! { "session_id": &session_id })
+                    .await;
+            }
+        }
+    }
+    let _ = db
+        .pentest_sessions()
+        .delete_many(doc! { "target_id": target_id })
+        .await;
+
+    // Delete DAST scan runs and their findings
+    let _ = db
+        .dast_findings()
+        .delete_many(doc! { "target_id": target_id })
+        .await;
+    let _ = db
+        .dast_scan_runs()
+        .delete_many(doc! { "target_id": target_id })
+        .await;
+
+    // Delete the target itself
+    if let Ok(oid) = mongodb::bson::oid::ObjectId::parse_str(target_id) {
+        let _ = db.dast_targets().delete_one(doc! { "_id": oid }).await;
+    }
+}

compliance-agent/src/api/handlers/sbom.rs

@@ -6,6 +6,7 @@ use mongodb::bson::doc;
 
 use super::dto::*;
 use compliance_core::models::SbomEntry;
+use compliance_core::tenant_ctx::TenantCtx;
 
 const COPYLEFT_LICENSES: &[&str] = &[
     "GPL-2.0",
@@ -29,8 +30,10 @@ const COPYLEFT_LICENSES: &[&str] = &[
 #[tracing::instrument(skip_all)]
 pub async fn sbom_filters(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
 ) -> Result<Json<serde_json::Value>, StatusCode> {
-    let db = &agent.db;
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
 
     let managers: Vec<String> = db
         .sbom_entries()
@@ -61,9 +64,11 @@ pub async fn sbom_filters(
 #[tracing::instrument(skip_all, fields(repo_id = ?filter.repo_id, package_manager = ?filter.package_manager))]
 pub async fn list_sbom(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Query(filter): Query<SbomFilter>,
 ) -> ApiResult<Vec<SbomEntry>> {
-    let db = &agent.db;
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
     let mut query = doc! {};
 
     if let Some(repo_id) = &filter.repo_id {
@@ -120,9 +125,11 @@ pub async fn list_sbom(
 #[tracing::instrument(skip_all)]
 pub async fn export_sbom(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Query(params): Query<SbomExportParams>,
 ) -> Result<impl IntoResponse, StatusCode> {
-    let db = &agent.db;
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
     let entries: Vec<SbomEntry> = match db
         .sbom_entries()
         .find(doc! { "repo_id": &params.repo_id })
@@ -236,9 +243,11 @@ pub async fn export_sbom(
 #[tracing::instrument(skip_all)]
 pub async fn license_summary(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Query(params): Query<SbomFilter>,
 ) -> ApiResult<Vec<LicenseSummary>> {
-    let db = &agent.db;
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
     let mut query = doc! {};
     if let Some(repo_id) = &params.repo_id {
         query.insert("repo_id", repo_id);
@@ -285,9 +294,11 @@ pub async fn license_summary(
 #[tracing::instrument(skip_all)]
 pub async fn sbom_diff(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Query(params): Query<SbomDiffParams>,
 ) -> ApiResult<SbomDiffResult> {
-    let db = &agent.db;
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
 
     let entries_a: Vec<SbomEntry> = match db
         .sbom_entries()

compliance-agent/src/api/handlers/scans.rs

@@ -4,13 +4,16 @@ use mongodb::bson::doc;
 
 use super::dto::*;
 use compliance_core::models::ScanRun;
+use compliance_core::tenant_ctx::TenantCtx;
 
 #[tracing::instrument(skip_all)]
 pub async fn list_scan_runs(
     Extension(agent): AgentExt,
+    tenant: TenantCtx,
     Query(params): Query<PaginationParams>,
 ) -> ApiResult<Vec<ScanRun>> {
-    let db = &agent.db;
+    let db = tenant_db(&agent, &tenant).await?;
+    let db = &db;
     let skip = (params.page.saturating_sub(1)) * params.limit as u64;
     let total = db.scan_runs().count_documents(doc! {}).await.unwrap_or(0);
 

compliance-agent/src/api/mod.rs

@@ -1,4 +1,3 @@
-pub mod auth_middleware;
 pub mod handlers;
 pub mod routes;
 pub mod server;

compliance-agent/src/api/routes.rs

@@ -47,6 +47,15 @@ pub fn build_router() -> Router {
         .route("/api/v1/sbom/diff", get(handlers::sbom_diff))
         .route("/api/v1/issues", get(handlers::list_issues))
         .route("/api/v1/scan-runs", get(handlers::list_scan_runs))
+        // MCP token management (per-tenant API tokens for the MCP server)
+        .route(
+            "/api/v1/mcp-tokens",
+            get(handlers::mcp_tokens::list_mcp_tokens).post(handlers::mcp_tokens::create_mcp_token),
+        )
+        .route(
+            "/api/v1/mcp-tokens/{id}",
+            delete(handlers::mcp_tokens::revoke_mcp_token),
+        )
         // Graph API endpoints
         .route("/api/v1/graph/{repo_id}", get(handlers::graph::get_graph))
         .route(
@@ -99,6 +108,29 @@ pub fn build_router() -> Router {
             "/api/v1/chat/{repo_id}/status",
             get(handlers::chat::embedding_status),
         )
+        // Help chat (documentation-grounded Q&A)
+        .route("/api/v1/help/chat", post(handlers::help_chat::help_chat))
+        // CVE notification endpoints
+        .route(
+            "/api/v1/notifications",
+            get(handlers::notifications::list_notifications),
+        )
+        .route(
+            "/api/v1/notifications/count",
+            get(handlers::notifications::notification_count),
+        )
+        .route(
+            "/api/v1/notifications/read-all",
+            post(handlers::notifications::mark_all_read),
+        )
+        .route(
+            "/api/v1/notifications/{id}/read",
+            patch(handlers::notifications::mark_read),
+        )
+        .route(
+            "/api/v1/notifications/{id}/dismiss",
+            patch(handlers::notifications::dismiss_notification),
+        )
         // Pentest API endpoints
         .route(
             "/api/v1/pentest/lookup-repo",

compliance-agent/src/api/server.rs

@@ -1,20 +1,76 @@
 use std::sync::Arc;
 
+use axum::extract::Request;
+use axum::http::HeaderValue;
+use axum::middleware::Next;
+use axum::response::Response;
 use axum::{middleware, Extension};
 use tokio::sync::RwLock;
 use tower_http::cors::CorsLayer;
+use tower_http::set_header::SetResponseHeaderLayer;
 use tower_http::trace::TraceLayer;
 
+use compliance_core::auth::{require_jwt_auth, require_tenant_status, JwksState};
+use compliance_core::{TenantContext, TenantStatus};
+
 use crate::agent::ComplianceAgent;
-use crate::api::auth_middleware::{require_jwt_auth, JwksState};
 use crate::api::routes;
 use crate::error::AgentError;
 
+/// Synthetic tenant id used when Keycloak isn't configured (local dev,
+/// `cargo run` against a bare Mongo). Lets the handler stack stay
+/// uniformly tenant-scoped without the operator having to spin up KC
+/// just to poke at the API. Override via `DEV_TENANT_ID`.
+const DEFAULT_DEV_TENANT_ID: &str = "dev";
+
+/// Inject a synthetic [`TenantContext`] for any request that lacks one.
+/// Only mounted when Keycloak is NOT configured; with KC, the real
+/// `require_jwt_auth` middleware owns this and we never reach here
+/// without a context.
+///
+/// Public so the integration-test harness can mount it without
+/// duplicating the synthetic-context shape.
+pub async fn inject_dev_tenant(mut request: Request, next: Next) -> Response {
+    if request.extensions().get::<TenantContext>().is_none() {
+        let tenant_id =
+            std::env::var("DEV_TENANT_ID").unwrap_or_else(|_| DEFAULT_DEV_TENANT_ID.to_string());
+        let ctx = TenantContext {
+            tenant_slug: tenant_id.clone(),
+            tenant_id,
+            org_roles: vec![],
+            products: vec![],
+            plan: "dev".to_string(),
+            status: TenantStatus::Active,
+            user_id: "dev-user".to_string(),
+            user_name: None,
+        };
+        request.extensions_mut().insert(ctx);
+    }
+    next.run(request).await
+}
+
 pub async fn start_api_server(agent: ComplianceAgent, port: u16) -> Result<(), AgentError> {
     let mut app = routes::build_router()
         .layer(Extension(Arc::new(agent.clone())))
         .layer(CorsLayer::permissive())
-        .layer(TraceLayer::new_for_http());
+        .layer(TraceLayer::new_for_http())
+        // Security headers (defense-in-depth, primary enforcement via Traefik)
+        .layer(SetResponseHeaderLayer::overriding(
+            axum::http::header::STRICT_TRANSPORT_SECURITY,
+            HeaderValue::from_static("max-age=31536000; includeSubDomains"),
+        ))
+        .layer(SetResponseHeaderLayer::overriding(
+            axum::http::header::X_FRAME_OPTIONS,
+            HeaderValue::from_static("DENY"),
+        ))
+        .layer(SetResponseHeaderLayer::overriding(
+            axum::http::header::X_CONTENT_TYPE_OPTIONS,
+            HeaderValue::from_static("nosniff"),
+        ))
+        .layer(SetResponseHeaderLayer::overriding(
+            axum::http::header::REFERRER_POLICY,
+            HeaderValue::from_static("strict-origin-when-cross-origin"),
+        ));
 
     if let (Some(kc_url), Some(kc_realm)) =
         (&agent.config.keycloak_url, &agent.config.keycloak_realm)
@@ -25,11 +81,22 @@ pub async fn start_api_server(agent: ComplianceAgent, port: u16) -> Result<(), A
             jwks_url,
         };
         tracing::info!("Keycloak JWT auth enabled for realm '{kc_realm}'");
+        // Layers execute outermost-first. Extension(jwks_state) must run
+        // before require_jwt_auth so the middleware can read it; the
+        // status gate runs after JWT so TenantContext is in extensions.
         app = app
-            .layer(Extension(jwks_state))
-            .layer(middleware::from_fn(require_jwt_auth));
+            .layer(middleware::from_fn(require_tenant_status))
+            .layer(middleware::from_fn(require_jwt_auth))
+            .layer(Extension(jwks_state));
     } else {
-        tracing::warn!("Keycloak not configured - API endpoints are unprotected");
+        let tenant_id =
+            std::env::var("DEV_TENANT_ID").unwrap_or_else(|_| DEFAULT_DEV_TENANT_ID.to_string());
+        tracing::warn!(
+            tenant_id = %tenant_id,
+            "Keycloak not configured — running unauthenticated against the dev tenant. \
+             DO NOT use in any environment with real customer data."
+        );
+        app = app.layer(middleware::from_fn(inject_dev_tenant));
     }
 
     let addr = format!("0.0.0.0:{port}");

compliance-agent/src/config.rs

@@ -42,7 +42,7 @@ pub fn load_config() -> Result<AgentConfig, AgentError> {
             .unwrap_or(3001),
         scan_schedule: env_var_opt("SCAN_SCHEDULE").unwrap_or_else(|| "0 0 */6 * * *".to_string()),
         cve_monitor_schedule: env_var_opt("CVE_MONITOR_SCHEDULE")
-            .unwrap_or_else(|| "0 0 0 * * *".to_string()),
+            .unwrap_or_else(|| "0 0 * * * *".to_string()),
         git_clone_base_path: env_var_opt("GIT_CLONE_BASE_PATH")
             .unwrap_or_else(|| "/tmp/compliance-scanner/repos".to_string()),
         ssh_key_path: env_var_opt("SSH_KEY_PATH")
@@ -54,6 +54,9 @@ pub fn load_config() -> Result<AgentConfig, AgentError> {
         pentest_verification_email: env_var_opt("PENTEST_VERIFICATION_EMAIL"),
         pentest_imap_host: env_var_opt("PENTEST_IMAP_HOST"),
         pentest_imap_port: env_var_opt("PENTEST_IMAP_PORT").and_then(|p| p.parse().ok()),
+        pentest_imap_tls: env_var_opt("PENTEST_IMAP_TLS")
+            .map(|v| v == "1" || v.eq_ignore_ascii_case("true"))
+            .unwrap_or(true),
         pentest_imap_username: env_var_opt("PENTEST_IMAP_USERNAME"),
         pentest_imap_password: env_secret_opt("PENTEST_IMAP_PASSWORD"),
     })

compliance-agent/src/database.rs

@@ -1,11 +1,216 @@
+use std::sync::Arc;
+
+use dashmap::DashMap;
 use mongodb::bson::doc;
 use mongodb::options::IndexOptions;
 use mongodb::{Client, Collection, IndexModel};
+use sha2::{Digest, Sha256};
 
 use compliance_core::models::*;
+use compliance_core::TenantContext;
 
 use crate::error::AgentError;
 
+/// Mongo enforces a 63-byte cap on database names (older clusters: 64
+/// on Linux, 63 on Windows; we target the conservative limit).
+const MAX_DB_NAME_LEN: usize = 63;
+
+/// Hex length of the SHA-256 truncation used for the hash fallback
+/// tenant DB name (16 bytes → 32 hex chars). 16 bytes gives ~2^64
+/// birthday-collision resistance — at our 10s-100s tenant scale this
+/// is effectively impossible to hit.
+const HASH_HEX_LEN: usize = 32;
+
+/// Largest `db_prefix` that still guarantees the hash-fallback name
+/// fits in the 63-byte cap: `prefix + "_" + 32 hex chars`.
+const MAX_PREFIX_LEN: usize = MAX_DB_NAME_LEN - 1 - HASH_HEX_LEN;
+
+/// Per-tenant Mongo connection broker (M7.2 isolation model).
+///
+/// Holds one [`Client`] and hands out [`Database`] handles physically
+/// scoped to `<db_prefix>_<tenant_id>`. The driver is the isolation
+/// boundary — a handle for tenant A cannot see tenant B's documents
+/// because it is connected to a different database, not because of an
+/// application-level filter.
+///
+/// Index creation runs idempotently the first time each tenant is seen
+/// in the process's lifetime. Mongo's `createIndex` is itself idempotent
+/// by index name; the in-memory `ensured` set just skips the round-trip.
+#[derive(Clone, Debug)]
+pub struct DatabasePool {
+    client: Client,
+    db_prefix: String,
+    ensured: Arc<DashMap<String, ()>>,
+}
+
+impl DatabasePool {
+    /// Connect to the cluster and prepare to hand out tenant databases
+    /// named `<db_prefix>_<tenant_id>`.
+    ///
+    /// Validates `db_prefix.len() <= MAX_PREFIX_LEN` so the
+    /// hash-fallback path is provably within Mongo's 63-byte db-name
+    /// cap. Refuses to construct a pool that could ever produce an
+    /// over-long name.
+    pub async fn connect(uri: &str, db_prefix: &str) -> Result<Self, AgentError> {
+        if db_prefix.len() > MAX_PREFIX_LEN {
+            return Err(AgentError::Other(format!(
+                "db_prefix '{db_prefix}' is {} chars; max is {MAX_PREFIX_LEN} so the \
+                 hash-fallback tenant DB name fits Mongo's {MAX_DB_NAME_LEN}-byte cap",
+                db_prefix.len()
+            )));
+        }
+        let client = Client::with_uri_str(uri).await?;
+        client
+            .database("admin")
+            .run_command(doc! { "ping": 1 })
+            .await?;
+        tracing::info!(
+            "MongoDB cluster reachable; per-tenant pool ready (db prefix '{db_prefix}')"
+        );
+        Ok(Self {
+            client,
+            db_prefix: db_prefix.to_string(),
+            ensured: Arc::new(DashMap::new()),
+        })
+    }
+
+    /// Return a [`Database`] scoped to this tenant. Ensures indexes on
+    /// first call per tenant (per process). Cheap on the hot path —
+    /// subsequent calls skip the round-trip.
+    pub async fn for_tenant(&self, ctx: &TenantContext) -> Result<Database, AgentError> {
+        self.for_tenant_id(&ctx.tenant_id).await
+    }
+
+    /// Like [`Self::for_tenant`] but accepts a bare tenant_id.
+    /// For background paths (scheduler, webhooks, pipeline orchestrators)
+    /// that don't have a full [`TenantContext`] but know which tenant
+    /// they're operating on (typically resolved from a URL path, a job
+    /// argument, or the registry).
+    pub async fn for_tenant_id(&self, tenant_id: &str) -> Result<Database, AgentError> {
+        let db_name = self.tenant_db_name(tenant_id);
+        let db = Database::from_database(self.client.database(&db_name));
+        // `DashMap::insert` returns the previous value; `None` means we
+        // were the first writer for this tenant_id and own the
+        // index-ensure work.
+        if self.ensured.insert(tenant_id.to_string(), ()).is_none() {
+            if let Err(e) = db.ensure_indexes().await {
+                // Roll the marker back so the next request retries.
+                self.ensured.remove(tenant_id);
+                return Err(e);
+            }
+            tracing::debug!(
+                tenant_id = %tenant_id,
+                db_name = %db_name,
+                "Indexes ensured for tenant database"
+            );
+        }
+        Ok(db)
+    }
+
+    /// Compute the Mongo database name for a tenant. Public for tests
+    /// and tenant offboarding (`pool.client().database(name).drop()`).
+    ///
+    /// Format: `<prefix>_<sanitized_tenant_id>` if it fits the 63-byte
+    /// cap, else `<prefix>_<sha256-16-byte-hex-of-tenant_id>`. The
+    /// `db_prefix` length invariant established at [`Self::connect`]
+    /// guarantees the hash-fallback name always fits — no runtime
+    /// assertion needed.
+    ///
+    /// Collision resistance: the hash fallback is a 16-byte SHA-256
+    /// truncation, which gives ~2^64 birthday-collision resistance. At
+    /// our 10s–100s tenant scale the probability of two tenant_ids
+    /// colliding is effectively zero. (8-byte truncation would have
+    /// been ~2^32 — too close for comfort on a regulated product.)
+    pub fn tenant_db_name(&self, tenant_id: &str) -> String {
+        let sanitized = sanitize_tenant_id(tenant_id);
+        let natural = format!("{}_{}", self.db_prefix, sanitized);
+        if natural.len() <= MAX_DB_NAME_LEN {
+            natural
+        } else {
+            let mut hasher = Sha256::new();
+            hasher.update(tenant_id.as_bytes());
+            let digest = hasher.finalize();
+            let suffix = hex::encode(&digest[..HASH_HEX_LEN / 2]);
+            format!("{}_{}", self.db_prefix, suffix)
+        }
+    }
+
+    /// Raw client handle. Reserved for cross-tenant admin flows that
+    /// must opt in explicitly (tenant listing, drop-on-offboard).
+    pub fn client(&self) -> &Client {
+        &self.client
+    }
+
+    /// Cross-tenant admin database used by features that intentionally
+    /// span tenants (today: MCP bearer tokens — each token row carries
+    /// a `tenant_id` and the MCP server reads them to route requests).
+    ///
+    /// The name `<db_prefix>__admin` (double underscore) is reserved —
+    /// the sanitizer never produces it for a normal tenant DB because
+    /// the natural format is `<db_prefix>_<sanitized_tenant_id>` (one
+    /// underscore) and tenant_ids would have to start with `_admin` to
+    /// collide. New tenant provisioning should reject such ids.
+    pub fn admin_db(&self) -> mongodb::Database {
+        self.client.database(&self.admin_db_name())
+    }
+
+    /// Name of the admin database — public so tests / operators can
+    /// drop it via the raw client.
+    pub fn admin_db_name(&self) -> String {
+        format!("{}__admin", self.db_prefix)
+    }
+
+    /// List every Mongo database currently belonging to this pool,
+    /// identified by the `<db_prefix>_` prefix. The result is the raw
+    /// database names — opening one for offboarding/cleanup goes
+    /// through [`Self::client`].
+    ///
+    /// Note: hashed-fallback names (very long tenant_ids) lose the
+    /// original tenant_id at the cluster level — we know a database
+    /// exists for *some* tenant but not which one. In practice
+    /// tenant_ids are UUIDs (36 chars) and never hit the fallback,
+    /// so this is a theoretical concern, not an operational one.
+    pub async fn list_tenant_db_names(&self) -> Result<Vec<String>, AgentError> {
+        let prefix = format!("{}_", self.db_prefix);
+        let names = self.client.list_database_names().await?;
+        Ok(names
+            .into_iter()
+            .filter(|n| n.starts_with(&prefix))
+            .collect())
+    }
+
+    /// Drop the database for a specific tenant. Used by GDPR delete
+    /// and tenant offboarding. Idempotent — dropping a non-existent
+    /// database is a no-op at the driver level.
+    ///
+    /// Also evicts the tenant from the in-memory `ensured` set so a
+    /// later re-provision triggers fresh `ensure_indexes`.
+    pub async fn drop_tenant(&self, tenant_id: &str) -> Result<(), AgentError> {
+        let db_name = self.tenant_db_name(tenant_id);
+        self.client.database(&db_name).drop().await?;
+        self.ensured.remove(tenant_id);
+        tracing::info!(
+            tenant_id = %tenant_id,
+            db_name = %db_name,
+            "Dropped tenant database"
+        );
+        Ok(())
+    }
+}
+
+/// Mongo database names disallow `/`, `\`, `.`, `"`, `$`, ` `, and NUL.
+/// breakpilot-dev tenant_ids are UUIDs so this is belt-and-braces, but
+/// it lets the pool tolerate any future tenant_id shape without surprise.
+fn sanitize_tenant_id(tenant_id: &str) -> String {
+    tenant_id
+        .chars()
+        .map(|c| match c {
+            '/' | '\\' | '.' | '"' | '$' | ' ' | '\0' => '_',
+            c => c,
+        })
+        .collect()
+}
+
 #[derive(Clone, Debug)]
 pub struct Database {
     inner: mongodb::Database,
@@ -20,6 +225,12 @@ impl Database {
         Ok(Self { inner: db })
     }
 
+    /// Wrap an already-resolved Mongo database. Used by [`DatabasePool`]
+    /// to hand out tenant-scoped handles without a fresh client per tenant.
+    pub(crate) fn from_database(inner: mongodb::Database) -> Self {
+        Self { inner }
+    }
+
     pub async fn ensure_indexes(&self) -> Result<(), AgentError> {
         // repositories: unique git_url
         self.repositories()
@@ -78,6 +289,25 @@ impl Database {
             )
             .await?;
 
+        // cve_notifications: unique cve_id + repo_id + package, status filter
+        self.cve_notifications()
+            .create_index(
+                IndexModel::builder()
+                    .keys(
+                        doc! { "cve_id": 1, "repo_id": 1, "package_name": 1, "package_version": 1 },
+                    )
+                    .options(IndexOptions::builder().unique(true).build())
+                    .build(),
+            )
+            .await?;
+        self.cve_notifications()
+            .create_index(
+                IndexModel::builder()
+                    .keys(doc! { "status": 1, "created_at": -1 })
+                    .build(),
+            )
+            .await?;
+
         // tracker_issues: unique finding_id
         self.tracker_issues()
             .create_index(
@@ -222,6 +452,12 @@ impl Database {
         self.inner.collection("cve_alerts")
     }
 
+    pub fn cve_notifications(
+        &self,
+    ) -> Collection<compliance_core::models::notification::CveNotification> {
+        self.inner.collection("cve_notifications")
+    }
+
     pub fn tracker_issues(&self) -> Collection<TrackerIssue> {
         self.inner.collection("tracker_issues")
     }

compliance-agent/src/lib.rs

@@ -0,0 +1,16 @@
+// Library entrypoint — re-exports for integration tests and the binary.
+
+pub mod agent;
+pub mod api;
+pub mod config;
+pub mod database;
+pub mod error;
+pub mod llm;
+pub mod pentest;
+pub mod pipeline;
+pub mod rag;
+pub mod scheduler;
+pub mod ssh;
+#[allow(dead_code)]
+pub mod trackers;
+pub mod webhooks;

compliance-agent/src/llm/client.rs

@@ -19,12 +19,17 @@ impl LlmClient {
         model: String,
         embed_model: String,
     ) -> Self {
+        let http = reqwest::Client::builder()
+            .timeout(std::time::Duration::from_secs(300))
+            .connect_timeout(std::time::Duration::from_secs(10))
+            .build()
+            .unwrap_or_default();
         Self {
             base_url,
             api_key,
             model,
             embed_model,
-            http: reqwest::Client::new(),
+            http,
         }
     }
 

compliance-agent/src/llm/descriptions.rs

@@ -5,15 +5,20 @@ use compliance_core::models::Finding;
 use crate::error::AgentError;
 use crate::llm::LlmClient;
 
-const DESCRIPTION_SYSTEM_PROMPT: &str = r#"You are a security engineer writing issue descriptions for a bug tracker. Generate a clear, actionable issue body in Markdown format that includes:
+const DESCRIPTION_SYSTEM_PROMPT: &str = r#"You are a security engineer writing a bug tracker issue for a developer to fix. Be direct and actionable — developers skim issue descriptions, so lead with what matters.
 
-1. **Summary**: 1-2 sentence overview
-2. **Evidence**: Code location, snippet, and what was detected
-3. **Impact**: What could happen if not fixed
-4. **Remediation**: Step-by-step fix instructions
-5. **References**: Relevant CWE/CVE links if applicable
+Format in Markdown:
 
-Keep it concise and professional. Use code blocks for code snippets."#;
+1. **What**: 1 sentence — what's wrong and where (file:line)
+2. **Why it matters**: 1-2 sentences — concrete impact if not fixed. Avoid generic "could lead to" phrasing; describe the specific attack or failure scenario.
+3. **Fix**: The specific code change needed. Use a code block with the corrected code if possible. If the fix is configuration-based, show the exact config change.
+4. **References**: CWE/CVE link if applicable (one line, not a section)
+
+Rules:
+- No filler paragraphs or background explanations
+- No restating the finding title in the body
+- Code blocks should show the FIX, not the vulnerable code (the developer can see that in the diff)
+- If the remediation is a one-liner, just say it — don't wrap it in a section header"#;
 
 pub async fn generate_issue_description(
     llm: &Arc<LlmClient>,

compliance-agent/src/llm/fixes.rs

@@ -5,7 +5,24 @@ use compliance_core::models::Finding;
 use crate::error::AgentError;
 use crate::llm::LlmClient;
 
-const FIX_SYSTEM_PROMPT: &str = r#"You are a security engineer. Given a security finding with code context, suggest a concrete code fix. Return ONLY the fixed code snippet that can directly replace the vulnerable code. Include brief inline comments explaining the fix."#;
+const FIX_SYSTEM_PROMPT: &str = r#"You are a security engineer suggesting a code fix. Return ONLY the corrected code that replaces the vulnerable snippet — no explanations, no markdown fences, no before/after comparison.
+
+Rules:
+- The fix must be a drop-in replacement for the vulnerable code
+- Preserve the original code's style, indentation, and naming conventions
+- Add at most one brief inline comment on the changed line explaining the security fix
+- If the fix requires importing a new module, include the import on a separate line prefixed with the language's comment syntax + "Add import: "
+- Do not refactor, rename variables, or "improve" unrelated code
+- If the vulnerability is a false positive and the code is actually safe, return the original code unchanged with a comment explaining why no fix is needed
+
+Language-specific fix guidance:
+- Rust: use `?` for error propagation, prefer `SecretString` for secrets, use parameterized queries with `sqlx`/`diesel`
+- Python: use parameterized queries (never f-strings in SQL), use `secrets` module not `random`, use `subprocess.run([...])` list form, use `markupsafe.escape()` for HTML
+- Go: use `sql.Query` with `$1`/`?` placeholders, use `crypto/rand` not `math/rand`, use `html/template` not `text/template`, return errors don't panic
+- Java/Kotlin: use `PreparedStatement` with `?` params, use `SecureRandom`, use `Jsoup.clean()` for HTML sanitization, use `@Valid` for input validation
+- Ruby: use ActiveRecord parameterized finders, use `SecureRandom`, use `ERB::Util.html_escape`, use `strong_parameters`
+- PHP: use PDO prepared statements with `:param` or `?`, use `random_bytes()`/`random_int()`, use `htmlspecialchars()` with `ENT_QUOTES`, use `password_hash(PASSWORD_BCRYPT)`
+- C/C++: use `snprintf` not `sprintf`, use bounds-checked APIs, free resources in reverse allocation order, use `memset_s` for secret cleanup"#;
 
 pub async fn suggest_fix(llm: &Arc<LlmClient>, finding: &Finding) -> Result<String, AgentError> {
     let user_prompt = format!(

compliance-agent/src/llm/review_prompts.rs

@@ -1,69 +1,138 @@
 // System prompts for multi-pass LLM code review.
 // Each pass focuses on a different aspect to avoid overloading a single prompt.
 
-pub const LOGIC_REVIEW_PROMPT: &str = r#"You are a senior software engineer reviewing code changes. Focus ONLY on logic and correctness issues.
+pub const LOGIC_REVIEW_PROMPT: &str = r#"You are a senior software engineer reviewing a code diff. Report ONLY genuine logic bugs that would cause incorrect behavior at runtime.
 
-Look for:
-- Off-by-one errors, wrong comparisons, missing edge cases
-- Incorrect control flow (unreachable code, missing returns, wrong loop conditions)
-- Race conditions or concurrency bugs
-- Resource leaks (unclosed handles, missing cleanup)
-- Wrong variable used (copy-paste errors)
-- Incorrect error handling (swallowed errors, wrong error type)
+Report:
+- Off-by-one errors, wrong comparisons, missing edge cases that cause wrong results
+- Incorrect control flow that produces wrong output (not style preferences)
+- Actual race conditions with concrete shared-state mutation (not theoretical ones)
+- Resource leaks where cleanup is truly missing (not just "could be improved")
+- Wrong variable used (copy-paste errors) — must be provably wrong, not just suspicious
+- Swallowed errors that silently hide failures in a way that matters
 
-Ignore: style, naming, formatting, documentation, minor improvements.
+Do NOT report:
+- Style, naming, formatting, documentation, or code organization preferences
+- Theoretical issues without a concrete triggering scenario
+- "Potential" problems that require assumptions not supported by the visible code
+- Complexity or function length — that's a separate review pass
 
-For each issue found, respond with a JSON array:
+Language-idiomatic patterns that are NOT bugs (do not flag these):
+- Rust: `||`/`&&` short-circuit evaluation, variable shadowing, `let` rebinding, `clone()`, `impl` blocks, `match` arms with guards, `?` operator chaining, `unsafe` blocks with safety comments
+- Python: duck typing, EAFP pattern (try/except vs check-first), `*args`/`**kwargs`, walrus operator `:=`, truthiness checks on containers, bare `except:` in top-level handlers
+- Go: multiple return values for errors, `if err != nil` patterns, goroutine + channel patterns, blank identifier `_`, named returns, `defer` for cleanup, `init()` functions
+- Java/Kotlin: checked exception patterns, method overloading, `Optional` vs null checks, Kotlin `?.` safe calls, `!!` non-null assertions in tests, `when` exhaustive matching, companion objects, `lateinit`
+- Ruby: monkey patching in libraries, method_missing, blocks/procs/lambdas, `rescue => e` patterns, `send`/`respond_to?` metaprogramming, `nil` checks via `&.` safe navigation
+- PHP: loose comparisons with `==` (only flag if `===` was clearly intended), `@` error suppression in legacy code, `isset()`/`empty()` patterns, magic methods (`__get`, `__call`), array functions as callbacks
+- C/C++: RAII patterns, move semantics, `const_cast`/`static_cast` in appropriate contexts, macro usage for platform compat, pointer arithmetic in low-level code, `goto` for cleanup in C
+
+Severity guide:
+- high: Will cause incorrect behavior in normal usage
+- medium: Will cause incorrect behavior in edge cases
+- low: Minor correctness concern with limited blast radius
+
+Prefer returning [] over reporting low-confidence guesses. A false positive wastes more developer time than a missed low-severity issue.
+
+Respond with a JSON array (no markdown fences):
 [{"title": "...", "description": "...", "severity": "high|medium|low", "file": "...", "line": N, "suggestion": "..."}]
 
 If no issues found, respond with: []"#;
 
-pub const SECURITY_REVIEW_PROMPT: &str = r#"You are a security engineer reviewing code changes. Focus ONLY on security vulnerabilities.
+pub const SECURITY_REVIEW_PROMPT: &str = r#"You are a security engineer reviewing a code diff. Report ONLY exploitable security vulnerabilities with a realistic attack scenario.
 
-Look for:
-- Injection vulnerabilities (SQL, command, XSS, template injection)
-- Authentication/authorization bypasses
-- Sensitive data exposure (logging secrets, hardcoded credentials)
-- Insecure cryptography (weak algorithms, predictable randomness)
-- Path traversal, SSRF, open redirects
-- Unsafe deserialization
-- Missing input validation at trust boundaries
+Report:
+- Injection vulnerabilities (SQL, command, XSS, template) where untrusted input reaches a sink
+- Authentication/authorization bypasses with a concrete exploit path
+- Sensitive data exposure: secrets in code, credentials in logs, PII leaks
+- Insecure cryptography: weak algorithms, predictable randomness, hardcoded keys
+- Path traversal, SSRF, open redirects — only where user input reaches the vulnerable API
+- Unsafe deserialization of untrusted data
+- Missing input validation at EXTERNAL trust boundaries (user input, API responses)
 
-Ignore: code style, performance, general quality.
+Do NOT report:
+- Internal code that only handles trusted/validated data
+- Hash functions used for non-security purposes (dedup fingerprints, cache keys, content addressing)
+- Logging of non-sensitive operational data (finding titles, counts, performance metrics)
+- "Information disclosure" for data that is already public or user-facing
+- Code style, performance, or general quality issues
+- Missing validation on internal function parameters (trust the caller within the same module/crate/package)
+- Theoretical attacks that require preconditions not present in the code
 
-For each issue found, respond with a JSON array:
+Language-specific patterns that are NOT vulnerabilities (do not flag these):
+- Python: `pickle` used on trusted internal data, `eval()`/`exec()` on hardcoded strings, `subprocess` with hardcoded commands, Django `mark_safe()` on static content, `assert` in non-security contexts
+- Go: `crypto/rand` is secure (don't confuse with `math/rand`), `sql.DB` with parameterized queries is safe, `http.ListenAndServe` without TLS in dev/internal, error strings in responses (Go convention)
+- Java/Kotlin: Spring Security annotations are sufficient auth checks, `@Transactional` provides atomicity, JPA parameterized queries are safe, Kotlin `require()`/`check()` are assertion patterns not vulnerabilities
+- Ruby: Rails `params.permit()` is input validation, `render html:` with `html_safe` on generated content, ActiveRecord parameterized finders are safe, Devise/Warden patterns for auth
+- PHP: PDO prepared statements are safe, Laravel Eloquent is parameterized, `htmlspecialchars()` is XSS mitigation, Symfony security voters are auth checks, `password_hash()`/`password_verify()` are correct bcrypt usage
+- C/C++: `strncpy`/`snprintf` are bounds-checked (vs `strcpy`/`sprintf`), smart pointers manage memory, RAII handles cleanup, `static_assert` is compile-time only, OpenSSL with proper context setup
+- Rust: `sha2`/`blake3` for fingerprinting is not "weak crypto", `unsafe` with documented invariants, `secrecy::SecretString` properly handles secrets
+
+Severity guide:
+- critical: Remote code execution, auth bypass, or data breach with no preconditions
+- high: Exploitable vulnerability requiring minimal preconditions
+- medium: Vulnerability requiring specific conditions or limited impact
+
+Prefer returning [] over reporting speculative vulnerabilities. Every false positive erodes trust in the scanner.
+
+Respond with a JSON array (no markdown fences):
 [{"title": "...", "description": "...", "severity": "critical|high|medium", "file": "...", "line": N, "cwe": "CWE-XXX", "suggestion": "..."}]
 
 If no issues found, respond with: []"#;
 
-pub const CONVENTION_REVIEW_PROMPT: &str = r#"You are a code reviewer checking adherence to project conventions. Focus ONLY on patterns that indicate likely bugs or maintenance problems.
+pub const CONVENTION_REVIEW_PROMPT: &str = r#"You are a code reviewer checking for convention violations that indicate likely bugs. Report ONLY deviations from the project's visible patterns that could cause real problems.
 
-Look for:
-- Inconsistent error handling patterns within the same module
-- Public API that doesn't follow the project's established patterns
-- Missing or incorrect type annotations that could cause runtime issues
-- Anti-patterns specific to the language (e.g. unwrap in Rust library code, any in TypeScript)
+Report:
+- Inconsistent error handling within the same module where the inconsistency could hide failures
+- Public API that breaks the module's established contract (not just different style)
+- Anti-patterns that are bugs in this language: e.g. `unwrap()` in Rust library code where the CI enforces `clippy::unwrap_used`, `any` defeating TypeScript's type system
 
-Do NOT report: minor style preferences, documentation gaps, formatting.
-Only report issues with HIGH confidence that they deviate from the visible codebase conventions.
+Do NOT report:
+- Style preferences, formatting, naming conventions, or documentation
+- Code organization suggestions ("this function should be split")
+- Patterns that are valid in the language even if you'd write them differently
+- "Missing type annotations" unless the code literally won't compile or causes a type inference bug
 
-For each issue found, respond with a JSON array:
+Language-specific patterns that are conventional (do not flag these):
+- Rust: variable shadowing, `||`/`&&` short-circuit, `let` rebinding, builder patterns, `clone()`, `From`/`Into` impl chains, `#[allow(...)]` attributes
+- Python: `**kwargs` forwarding, `@property` setters, `__dunder__` methods, list comprehensions with conditions, `if TYPE_CHECKING` imports, `noqa` comments
+- Go: stuttering names (`http.HTTPClient`) discouraged but not a bug, `context.Context` as first param, init() functions, `//nolint` directives, returning concrete types vs interfaces in internal code
+- Java/Kotlin: builder pattern boilerplate, Lombok annotations (`@Data`, `@Builder`), Kotlin data classes, `companion object` factories, `@Suppress` annotations, checked exception wrapping
+- Ruby: `attr_accessor` usage, `Enumerable` mixin patterns, `module_function`, `class << self` syntax, DSL blocks (Rake, RSpec, Sinatra routes)
+- PHP: `__construct` with property promotion, Laravel facades, static factory methods, nullable types with `?`, attribute syntax `#[...]`
+- C/C++: header guards vs `#pragma once`, forward declarations, `const` correctness patterns, template specialization, `auto` type deduction
+
+Severity guide:
+- medium: Convention violation that will likely cause a bug or maintenance problem
+- low: Convention violation that is a minor concern
+
+Return at most 3 findings. Prefer [] over marginal findings.
+
+Respond with a JSON array (no markdown fences):
 [{"title": "...", "description": "...", "severity": "medium|low", "file": "...", "line": N, "suggestion": "..."}]
 
 If no issues found, respond with: []"#;
 
-pub const COMPLEXITY_REVIEW_PROMPT: &str = r#"You are reviewing code changes for excessive complexity that could lead to bugs.
+pub const COMPLEXITY_REVIEW_PROMPT: &str = r#"You are reviewing code changes for complexity that is likely to cause bugs. Report ONLY complexity that makes the code demonstrably harder to reason about.
 
-Look for:
-- Functions over 50 lines that should be decomposed
-- Deeply nested control flow (4+ levels)
-- Complex boolean expressions that are hard to reason about
-- Functions with 5+ parameters
-- Code duplication within the changed files
+Report:
+- Functions over 80 lines with multiple interleaved responsibilities (not just long)
+- Deeply nested control flow (5+ levels) where flattening would prevent bugs
+- Complex boolean expressions that a reader would likely misinterpret
 
-Only report complexity issues that are HIGH risk for future bugs. Ignore acceptable complexity in configuration, CLI argument parsing, or generated code.
+Do NOT report:
+- Functions that are long but linear and easy to follow
+- Acceptable complexity: configuration setup, CLI parsing, test helpers, builder patterns
+- Code that is complex because the problem is complex — only report if restructuring would reduce bug risk
+- "This function does multiple things" unless you can identify a specific bug risk from the coupling
+- Suggestions that would just move complexity elsewhere without reducing it
 
-For each issue found, respond with a JSON array:
+Severity guide:
+- medium: Complexity that has a concrete risk of causing bugs during future changes
+- low: Complexity that makes review harder but is unlikely to cause bugs
+
+Return at most 2 findings. Prefer [] over reporting complexity that is justified.
+
+Respond with a JSON array (no markdown fences):
 [{"title": "...", "description": "...", "severity": "medium|low", "file": "...", "line": N, "suggestion": "..."}]
 
 If no issues found, respond with: []"#;

compliance-agent/src/llm/triage.rs

@@ -8,22 +8,46 @@ use crate::pipeline::orchestrator::GraphContext;
 /// Maximum number of findings to include in a single LLM triage call.
 const TRIAGE_CHUNK_SIZE: usize = 30;
 
-const TRIAGE_SYSTEM_PROMPT: &str = r#"You are a security finding triage expert. Analyze each of the following security findings with its code context and determine the appropriate action.
+const TRIAGE_SYSTEM_PROMPT: &str = r#"You are a pragmatic security triage expert. Your job is to filter out noise and keep only findings that a developer should actually fix. Be aggressive about dismissing false positives — a clean, high-signal list is more valuable than a comprehensive one.
 
 Actions:
-- "confirm": The finding is a true positive at the reported severity. Keep as-is.
-- "downgrade": The finding is real but over-reported. Lower severity recommended.
-- "upgrade": The finding is under-reported. Higher severity recommended.
-- "dismiss": The finding is a false positive. Should be removed.
+- "confirm": True positive with real impact. Keep severity as-is.
+- "downgrade": Real issue but over-reported severity. Lower it.
+- "upgrade": Under-reported — higher severity warranted.
+- "dismiss": False positive, not exploitable, or not actionable. Remove it.
 
-Consider:
-- Is the code in a test, example, or generated file? (lower confidence for test code)
-- Does the surrounding code context confirm or refute the finding?
-- Is the finding actionable by a developer?
-- Would a real attacker be able to exploit this?
+Dismiss when:
+- The scanner flagged a language idiom as a bug (see examples below)
+- The finding is in test/example/generated/vendored code
+- The "vulnerability" requires preconditions that don't exist in the code
+- The finding is about code style, complexity, or theoretical concerns rather than actual bugs
+- A hash function is used for non-security purposes (dedup, caching, content addressing)
+- Internal logging of non-sensitive operational data is flagged as "information disclosure"
+- The finding duplicates another finding already in the list
+- Framework-provided security is already in place (e.g. ORM parameterized queries, CSRF middleware, auth decorators)
 
-Respond with a JSON array, one entry per finding in the same order they were presented:
-[{"id": "<fingerprint>", "action": "confirm|downgrade|upgrade|dismiss", "confidence": 0-10, "rationale": "brief explanation", "remediation": "optional fix suggestion"}, ...]"#;
+Common false positive patterns by language (dismiss these):
+- Rust: short-circuit `||`/`&&`, variable shadowing, `clone()`, `unsafe` with safety docs, `sha2` for fingerprinting
+- Python: EAFP try/except, `subprocess` with hardcoded args, `pickle` on trusted data, Django `mark_safe` on static content
+- Go: `if err != nil` is not "swallowed error", `crypto/rand` is secure, returning errors is not "information disclosure"
+- Java/Kotlin: Spring Security annotations are valid auth, JPA parameterized queries are safe, Kotlin `!!` in tests is fine
+- Ruby: Rails `params.permit` is validation, ActiveRecord finders are parameterized, `html_safe` on generated content
+- PHP: PDO prepared statements are safe, Laravel Eloquent is parameterized, `htmlspecialchars` is XSS mitigation
+- C/C++: `strncpy`/`snprintf` are bounds-checked, smart pointers manage memory, RAII handles cleanup
+
+Confirm only when:
+- You can describe a concrete scenario where the bug manifests or the vulnerability is exploitable
+- The fix is actionable (developer can change specific code to resolve it)
+- The finding is in production code that handles external input or sensitive data
+
+Confidence scoring (0-10):
+- 8-10: Certain true positive with clear exploit/bug scenario
+- 5-7: Likely true positive, some assumptions required
+- 3-4: Uncertain, needs manual review
+- 0-2: Almost certainly a false positive
+
+Respond with a JSON array, one entry per finding in the same order presented (no markdown fences):
+[{"id": "<fingerprint>", "action": "confirm|downgrade|upgrade|dismiss", "confidence": 0-10, "rationale": "1-2 sentences", "remediation": "optional fix"}, ...]"#;
 
 pub async fn triage_findings(
     llm: &Arc<LlmClient>,

compliance-agent/src/main.rs

@@ -1,23 +1,10 @@
-mod agent;
-mod api;
-pub(crate) mod config;
-mod database;
-mod error;
-mod llm;
-mod pentest;
-mod pipeline;
-mod rag;
-mod scheduler;
-mod ssh;
-#[allow(dead_code)]
-mod trackers;
-mod webhooks;
+use compliance_agent::{agent, api, config, database, scheduler, ssh, webhooks};
 
 #[tokio::main]
 async fn main() -> Result<(), Box<dyn std::error::Error>> {
     match dotenvy::dotenv() {
         Ok(path) => eprintln!("[dotenv] Loaded from: {}", path.display()),
-        Err(e) => eprintln!("[dotenv] FAILED: {e}"),
+        Err(_) => eprintln!("[dotenv] No .env file found, using environment variables"),
     }
 
     let _telemetry_guard = compliance_core::telemetry::init_telemetry("compliance-agent");
@@ -38,10 +25,13 @@ async fn main() -> Result<(), Box<dyn std::error::Error>> {
     }
 
     tracing::info!("Connecting to MongoDB...");
-    let db = database::Database::connect(&config.mongodb_uri, &config.mongodb_database).await?;
-    db.ensure_indexes().await?;
+    // Per-tenant pool only — the agent has no shared "default" database
+    // after M7.2-D. `mongodb_database` is now the db-name prefix used
+    // for tenant databases (`<prefix>_<tenant_id>`).
+    let db_pool =
+        database::DatabasePool::connect(&config.mongodb_uri, &config.mongodb_database).await?;
 
-    let agent = agent::ComplianceAgent::new(config.clone(), db.clone());
+    let agent = agent::ComplianceAgent::new(config.clone(), db_pool);
 
     tracing::info!("Starting scheduler...");
     let scheduler_agent = agent.clone();

compliance-agent/src/pentest/cleanup.rs

@@ -336,6 +336,7 @@ mod tests {
             pentest_verification_email: None,
             pentest_imap_host: None,
             pentest_imap_port: None,
+            pentest_imap_tls: true,
             pentest_imap_username: None,
             pentest_imap_password: None,
         }

compliance-agent/src/pentest/orchestrator.rs

@@ -321,9 +321,38 @@ impl PentestOrchestrator {
                                     total_findings += findings_count;
 
                                     let mut finding_ids: Vec<String> = Vec::new();
-                                    for mut finding in result.findings {
+                                    // Dedup findings within this tool result before inserting
+                                    let deduped_findings =
+                                        crate::pipeline::dedup::dedup_dast_findings(
+                                            result.findings,
+                                        );
+                                    for mut finding in deduped_findings {
                                         finding.scan_run_id = session_id.clone();
                                         finding.session_id = Some(session_id.clone());
+
+                                        // Check for existing duplicate in this session
+                                        let fp = crate::pipeline::dedup::compute_dast_fingerprint(
+                                            &finding,
+                                        );
+                                        let existing = self
+                                            .db
+                                            .dast_findings()
+                                            .find_one(doc! {
+                                                "session_id": &session_id,
+                                                "title": &finding.title,
+                                                "endpoint": &finding.endpoint,
+                                                "method": &finding.method,
+                                            })
+                                            .await;
+                                        if matches!(existing, Ok(Some(_))) {
+                                            tracing::debug!(
+                                                "Skipping duplicate DAST finding: {} (fp={:.12})",
+                                                finding.title,
+                                                fp,
+                                            );
+                                            continue;
+                                        }
+
                                         let insert_result =
                                             self.db.dast_findings().insert_one(&finding).await;
                                         if let Ok(res) = &insert_result {

compliance-agent/src/pentest/prompt_builder.rs

@@ -314,6 +314,21 @@ impl PentestOrchestrator {
 - For SPA apps: a 200 HTTP status does NOT mean the page is accessible — check the actual
   page content with the browser tool to verify if it shows real data or a login redirect.
 
+## Finding Quality Rules
+- **Do not report the same issue twice.** If multiple tools detect the same missing header or
+  vulnerability on the same endpoint, report it ONCE with the most specific tool's output.
+  For example, if the recon tool and the header scanner both find missing HSTS, report it only
+  from the header scanner (more specific).
+- **Group related findings.** Missing security headers on the same endpoint are ONE finding
+  ("Missing security headers") listing all missing headers, not separate findings per header.
+- **Severity must match real impact:**
+  - critical/high: Exploitable vulnerability (you can demonstrate the exploit)
+  - medium: Real misconfiguration with security implications but not directly exploitable
+  - low: Best-practice recommendation, defense-in-depth, or informational
+- **Missing headers are medium at most** unless you can demonstrate a concrete exploit enabled
+  by the missing header (e.g., missing CSP + confirmed XSS = high for CSP finding).
+- Console.log in third-party/vendored JS (node_modules, minified libraries) is informational only.
+
 ## Important
 - This is an authorized penetration test. All testing is permitted within the target scope.
 - Respect the rate limit of {rate_limit} requests per second.

compliance-agent/src/pipeline/code_review.rs

@@ -66,8 +66,10 @@ impl CodeReviewScanner {
             }
         }
 
+        let deduped = dedup_cross_pass(all_findings);
+
         ScanOutput {
-            findings: all_findings,
+            findings: deduped,
             sbom_entries: Vec::new(),
         }
     }
@@ -184,3 +186,51 @@ struct ReviewIssue {
     #[serde(default)]
     suggestion: Option<String>,
 }
+
+/// Deduplicate findings across review passes.
+///
+/// Multiple passes often flag the same issue (e.g. SQL injection reported by
+/// logic, security, and convention passes). We group by file + nearby line +
+/// normalized title keywords and keep the highest-severity finding.
+fn dedup_cross_pass(findings: Vec<Finding>) -> Vec<Finding> {
+    use std::collections::HashMap;
+
+    // Build a dedup key: (file, line bucket, normalized title words)
+    fn dedup_key(f: &Finding) -> String {
+        let file = f.file_path.as_deref().unwrap_or("");
+        // Group lines within 3 of each other
+        let line_bucket = f.line_number.unwrap_or(0) / 4;
+        // Normalize: lowercase, keep only alphanumeric, sort words for order-independence
+        let title_lower = f.title.to_lowercase();
+        let mut words: Vec<&str> = title_lower
+            .split(|c: char| !c.is_alphanumeric())
+            .filter(|w| w.len() > 2)
+            .collect();
+        words.sort();
+        format!("{file}:{line_bucket}:{}", words.join(","))
+    }
+
+    let mut groups: HashMap<String, Finding> = HashMap::new();
+
+    for finding in findings {
+        let key = dedup_key(&finding);
+        groups
+            .entry(key)
+            .and_modify(|existing| {
+                // Keep the higher severity; on tie, keep the one with more detail
+                if finding.severity > existing.severity
+                    || (finding.severity == existing.severity
+                        && finding.description.len() > existing.description.len())
+                {
+                    *existing = finding.clone();
+                }
+                // Merge CWE if the existing one is missing it
+                if existing.cwe.is_none() {
+                    existing.cwe = finding.cwe.clone();
+                }
+            })
+            .or_insert(finding);
+    }
+
+    groups.into_values().collect()
+}

compliance-agent/src/pipeline/dedup.rs

@@ -1,5 +1,7 @@
 use sha2::{Digest, Sha256};
 
+use compliance_core::models::dast::DastFinding;
+
 pub fn compute_fingerprint(parts: &[&str]) -> String {
     let mut hasher = Sha256::new();
     for part in parts {
@@ -9,9 +11,209 @@ pub fn compute_fingerprint(parts: &[&str]) -> String {
     hex::encode(hasher.finalize())
 }
 
+/// Compute a dedup fingerprint for a DAST finding.
+///
+/// The key is derived from the *canonicalized* title (lowercased, domain names
+/// stripped, known synonyms resolved), endpoint, and HTTP method.  This lets us
+/// detect both exact duplicates (same tool reporting twice across passes) and
+/// semantic duplicates (e.g., `security_header_missing` "Missing HSTS header"
+/// vs `tls_misconfiguration` "Missing strict-transport-security header").
+pub fn compute_dast_fingerprint(f: &DastFinding) -> String {
+    let canon = canonicalize_dast_title(&f.title);
+    let endpoint = f.endpoint.to_lowercase().trim_end_matches('/').to_string();
+    let method = f.method.to_uppercase();
+    let param = f.parameter.as_deref().unwrap_or("");
+    compute_fingerprint(&[&canon, &endpoint, &method, param])
+}
+
+/// Canonicalize a DAST finding title for dedup purposes.
+///
+/// 1. Lowercase
+/// 2. Strip domain names / URLs (e.g. "for comp-dev.meghsakha.com")
+/// 3. Resolve known header synonyms (hsts ↔ strict-transport-security, etc.)
+/// 4. Strip extra whitespace
+fn canonicalize_dast_title(title: &str) -> String {
+    let mut s = title.to_lowercase();
+
+    // Strip "for <domain>" or "on <domain>" suffixes
+    // Pattern: "for <word.word...>" or "on <method> <url>"
+    if let Some(idx) = s.find(" for ") {
+        // Check if what follows looks like a domain or URL
+        let rest = &s[idx + 5..];
+        if rest.contains('.') || rest.starts_with("http") {
+            s.truncate(idx);
+        }
+    }
+    if let Some(idx) = s.find(" on ") {
+        let rest = &s[idx + 4..];
+        if rest.contains("http") || rest.contains('/') {
+            s.truncate(idx);
+        }
+    }
+
+    // Resolve known header synonyms
+    let synonyms: &[(&str, &str)] = &[
+        ("hsts", "strict-transport-security"),
+        ("csp", "content-security-policy"),
+        ("cors", "cross-origin-resource-sharing"),
+        ("xfo", "x-frame-options"),
+    ];
+    for &(short, canonical) in synonyms {
+        // Only replace whole words — check boundaries
+        if let Some(pos) = s.find(short) {
+            let before_ok = pos == 0 || !s.as_bytes()[pos - 1].is_ascii_alphanumeric();
+            let after_ok = pos + short.len() >= s.len()
+                || !s.as_bytes()[pos + short.len()].is_ascii_alphanumeric();
+            if before_ok && after_ok {
+                s = format!("{}{}{}", &s[..pos], canonical, &s[pos + short.len()..]);
+            }
+        }
+    }
+
+    // Collapse whitespace
+    s.split_whitespace().collect::<Vec<_>>().join(" ")
+}
+
+/// Deduplicate a list of DAST findings, merging evidence from duplicates.
+///
+/// Two-phase approach:
+/// 1. **Exact dedup** — group by canonicalized `(title, endpoint, method, parameter)`.
+///    Merge evidence arrays, keep the highest severity, preserve exploitable flag.
+/// 2. **CWE-based dedup** — within the same `(cwe, endpoint, method)` group, merge
+///    findings whose canonicalized titles resolve to the same subject (e.g., HSTS
+///    reported as both `security_header_missing` and `tls_misconfiguration`).
+pub fn dedup_dast_findings(findings: Vec<DastFinding>) -> Vec<DastFinding> {
+    use std::collections::HashMap;
+
+    if findings.len() <= 1 {
+        return findings;
+    }
+
+    // Phase 1: exact fingerprint dedup
+    let mut seen: HashMap<String, usize> = HashMap::new();
+    let mut deduped: Vec<DastFinding> = Vec::new();
+
+    for finding in findings {
+        let fp = compute_dast_fingerprint(&finding);
+
+        if let Some(&idx) = seen.get(&fp) {
+            // Merge into existing
+            merge_dast_finding(&mut deduped[idx], &finding);
+        } else {
+            seen.insert(fp, deduped.len());
+            deduped.push(finding);
+        }
+    }
+
+    let before = deduped.len();
+
+    // Phase 2: CWE-based related dedup
+    // Group by (cwe, endpoint_normalized, method) — only when CWE is present
+    let mut cwe_groups: HashMap<String, Vec<usize>> = HashMap::new();
+    for (i, f) in deduped.iter().enumerate() {
+        if let Some(ref cwe) = f.cwe {
+            let key = format!(
+                "{}|{}|{}",
+                cwe,
+                f.endpoint.to_lowercase().trim_end_matches('/'),
+                f.method.to_uppercase(),
+            );
+            cwe_groups.entry(key).or_default().push(i);
+        }
+    }
+
+    // For each CWE group with multiple findings, keep the one with highest severity
+    // and most evidence, merge the rest into it
+    let mut merge_map: HashMap<usize, Vec<usize>> = HashMap::new();
+    let mut remove_indices: Vec<usize> = Vec::new();
+
+    for indices in cwe_groups.values() {
+        if indices.len() <= 1 {
+            continue;
+        }
+        // Find the "primary" finding: highest severity, then most evidence, then longest description
+        let Some(&primary_idx) = indices.iter().max_by(|&&a, &&b| {
+            deduped[a]
+                .severity
+                .cmp(&deduped[b].severity)
+                .then_with(|| deduped[a].evidence.len().cmp(&deduped[b].evidence.len()))
+                .then_with(|| {
+                    deduped[a]
+                        .description
+                        .len()
+                        .cmp(&deduped[b].description.len())
+                })
+        }) else {
+            continue;
+        };
+
+        for &idx in indices {
+            if idx != primary_idx {
+                remove_indices.push(idx);
+                merge_map.entry(primary_idx).or_default().push(idx);
+            }
+        }
+    }
+
+    if !remove_indices.is_empty() {
+        remove_indices.sort_unstable();
+        remove_indices.dedup();
+
+        // Merge evidence
+        for (&primary, secondaries) in &merge_map {
+            let extra_evidence: Vec<_> = secondaries
+                .iter()
+                .flat_map(|&i| deduped[i].evidence.clone())
+                .collect();
+            let any_exploitable = secondaries.iter().any(|&i| deduped[i].exploitable);
+
+            deduped[primary].evidence.extend(extra_evidence);
+            if any_exploitable {
+                deduped[primary].exploitable = true;
+            }
+        }
+
+        // Remove merged findings (iterate in reverse to preserve indices)
+        for &idx in remove_indices.iter().rev() {
+            deduped.remove(idx);
+        }
+    }
+
+    let after = deduped.len();
+    if before != after {
+        tracing::debug!(
+            "DAST CWE-based dedup: {before} → {after} findings ({} merged)",
+            before - after
+        );
+    }
+
+    deduped
+}
+
+/// Merge a duplicate DAST finding into a primary one.
+fn merge_dast_finding(primary: &mut DastFinding, duplicate: &DastFinding) {
+    primary.evidence.extend(duplicate.evidence.clone());
+    if duplicate.severity > primary.severity {
+        primary.severity = duplicate.severity.clone();
+    }
+    if duplicate.exploitable {
+        primary.exploitable = true;
+    }
+    // Keep the longer/better description
+    if duplicate.description.len() > primary.description.len() {
+        primary.description.clone_from(&duplicate.description);
+    }
+    // Keep remediation if primary doesn't have one
+    if primary.remediation.is_none() && duplicate.remediation.is_some() {
+        primary.remediation.clone_from(&duplicate.remediation);
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;
+    use compliance_core::models::dast::DastVulnType;
+    use compliance_core::models::finding::Severity;
 
     #[test]
     fn fingerprint_is_deterministic() {
@@ -55,4 +257,159 @@ mod tests {
         let b = compute_fingerprint(&["a", "bc"]);
         assert_ne!(a, b);
     }
+
+    fn make_dast(title: &str, endpoint: &str, vuln_type: DastVulnType) -> DastFinding {
+        let mut f = DastFinding::new(
+            "run1".into(),
+            "target1".into(),
+            vuln_type,
+            title.into(),
+            format!("Description for {title}"),
+            Severity::Medium,
+            endpoint.into(),
+            "GET".into(),
+        );
+        f.cwe = Some("CWE-319".into());
+        f
+    }
+
+    #[test]
+    fn canonicalize_strips_domain_suffix() {
+        let canon = canonicalize_dast_title("Missing HSTS header for comp-dev.meghsakha.com");
+        assert!(!canon.contains("meghsakha"), "domain should be stripped");
+        assert!(
+            canon.contains("strict-transport-security"),
+            "hsts should be resolved: {canon}"
+        );
+    }
+
+    #[test]
+    fn canonicalize_resolves_synonyms() {
+        let a = canonicalize_dast_title("Missing HSTS header");
+        let b = canonicalize_dast_title("Missing strict-transport-security header");
+        assert_eq!(a, b);
+    }
+
+    #[test]
+    fn exact_dedup_merges_identical_findings() {
+        let f1 = make_dast(
+            "Missing strict-transport-security header",
+            "https://example.com",
+            DastVulnType::SecurityHeaderMissing,
+        );
+        let f2 = make_dast(
+            "Missing strict-transport-security header",
+            "https://example.com",
+            DastVulnType::SecurityHeaderMissing,
+        );
+        let result = dedup_dast_findings(vec![f1, f2]);
+        assert_eq!(result.len(), 1, "exact duplicates should be merged");
+    }
+
+    #[test]
+    fn synonym_dedup_merges_hsts_variants() {
+        let f1 = make_dast(
+            "Missing strict-transport-security header",
+            "https://example.com",
+            DastVulnType::SecurityHeaderMissing,
+        );
+        let f2 = make_dast(
+            "Missing HSTS header for example.com",
+            "https://example.com",
+            DastVulnType::TlsMisconfiguration,
+        );
+        let result = dedup_dast_findings(vec![f1, f2]);
+        assert_eq!(
+            result.len(),
+            1,
+            "HSTS synonym variants should merge to 1 finding"
+        );
+    }
+
+    #[test]
+    fn different_headers_not_merged() {
+        let mut f1 = make_dast(
+            "Missing x-content-type-options header",
+            "https://example.com",
+            DastVulnType::SecurityHeaderMissing,
+        );
+        f1.cwe = Some("CWE-16".into());
+        let mut f2 = make_dast(
+            "Missing permissions-policy header",
+            "https://example.com",
+            DastVulnType::SecurityHeaderMissing,
+        );
+        f2.cwe = Some("CWE-16".into());
+        // These share CWE-16 but are different headers — phase 2 will merge them
+        // since they share the same CWE+endpoint. This is acceptable because they
+        // have the same root cause (missing security headers configuration).
+        let result = dedup_dast_findings(vec![f1, f2]);
+        // CWE-based dedup will merge these into 1
+        assert!(
+            result.len() <= 2,
+            "same CWE+endpoint findings may be merged"
+        );
+    }
+
+    #[test]
+    fn different_endpoints_not_merged() {
+        let f1 = make_dast(
+            "Missing strict-transport-security header",
+            "https://example.com",
+            DastVulnType::SecurityHeaderMissing,
+        );
+        let f2 = make_dast(
+            "Missing strict-transport-security header",
+            "https://other.com",
+            DastVulnType::SecurityHeaderMissing,
+        );
+        let result = dedup_dast_findings(vec![f1, f2]);
+        assert_eq!(result.len(), 2, "different endpoints should not merge");
+    }
+
+    #[test]
+    fn dedup_preserves_highest_severity() {
+        let f1 = make_dast(
+            "Missing strict-transport-security header",
+            "https://example.com",
+            DastVulnType::SecurityHeaderMissing,
+        );
+        let mut f2 = make_dast(
+            "Missing strict-transport-security header",
+            "https://example.com",
+            DastVulnType::SecurityHeaderMissing,
+        );
+        f2.severity = Severity::High;
+        let result = dedup_dast_findings(vec![f1, f2]);
+        assert_eq!(result.len(), 1);
+        assert_eq!(result[0].severity, Severity::High);
+    }
+
+    #[test]
+    fn dedup_merges_evidence() {
+        let mut f1 = make_dast(
+            "Missing strict-transport-security header",
+            "https://example.com",
+            DastVulnType::SecurityHeaderMissing,
+        );
+        f1.evidence
+            .push(compliance_core::models::dast::DastEvidence {
+                request_method: "GET".into(),
+                request_url: "https://example.com".into(),
+                request_headers: None,
+                request_body: None,
+                response_status: 200,
+                response_headers: None,
+                response_snippet: Some("pass 1".into()),
+                screenshot_path: None,
+                payload: None,
+                response_time_ms: None,
+            });
+        let mut f2 = f1.clone();
+        f2.evidence[0].response_snippet = Some("pass 2".into());
+
+        let result = dedup_dast_findings(vec![f1, f2]);
+        assert_eq!(result.len(), 1);
+        assert_eq!(result[0].evidence.len(), 2, "evidence should be merged");
+    }
 }

compliance-agent/src/pipeline/gitleaks.rs

@@ -19,26 +19,33 @@ impl Scanner for GitleaksScanner {
 
     #[tracing::instrument(skip_all)]
     async fn scan(&self, repo_path: &Path, repo_id: &str) -> Result<ScanOutput, CoreError> {
-        let output = tokio::process::Command::new("gitleaks")
-            .args([
-                "detect",
-                "--source",
-                ".",
-                "--report-format",
-                "json",
-                "--report-path",
-                "/dev/stdout",
-                "--no-banner",
-                "--exit-code",
-                "0",
-            ])
-            .current_dir(repo_path)
-            .output()
-            .await
-            .map_err(|e| CoreError::Scanner {
-                scanner: "gitleaks".to_string(),
-                source: Box::new(e),
-            })?;
+        let output = tokio::time::timeout(
+            std::time::Duration::from_secs(300),
+            tokio::process::Command::new("gitleaks")
+                .args([
+                    "detect",
+                    "--source",
+                    ".",
+                    "--report-format",
+                    "json",
+                    "--report-path",
+                    "/dev/stdout",
+                    "--no-banner",
+                    "--exit-code",
+                    "0",
+                ])
+                .current_dir(repo_path)
+                .output(),
+        )
+        .await
+        .map_err(|_| CoreError::Scanner {
+            scanner: "gitleaks".to_string(),
+            source: "timed out after 5 minutes".into(),
+        })?
+        .map_err(|e| CoreError::Scanner {
+            scanner: "gitleaks".to_string(),
+            source: Box::new(e),
+        })?;
 
         if output.stdout.is_empty() {
             return Ok(ScanOutput::default());

compliance-agent/src/pipeline/orchestrator.rs

@@ -10,7 +10,6 @@ use compliance_core::AgentConfig;
 use crate::database::Database;
 use crate::error::AgentError;
 use crate::llm::LlmClient;
-use crate::pipeline::code_review::CodeReviewScanner;
 use crate::pipeline::cve::CveScanner;
 use crate::pipeline::git::GitOps;
 use crate::pipeline::gitleaks::GitleaksScanner;
@@ -175,19 +174,26 @@ impl PipelineOrchestrator {
                 k.expose_secret().to_string()
             }),
         );
-        let cve_alerts = match async {
-            cve_scanner
-                .scan_dependencies(&repo_id, &mut sbom_entries)
-                .await
-        }
-        .instrument(tracing::info_span!("stage_cve_scanning"))
+        let cve_alerts = match tokio::time::timeout(
+            std::time::Duration::from_secs(600),
+            async {
+                cve_scanner
+                    .scan_dependencies(&repo_id, &mut sbom_entries)
+                    .await
+            }
+            .instrument(tracing::info_span!("stage_cve_scanning")),
+        )
         .await
         {
-            Ok(alerts) => alerts,
-            Err(e) => {
+            Ok(Ok(alerts)) => alerts,
+            Ok(Err(e)) => {
                 tracing::warn!("[{repo_id}] CVE scanning failed: {e}");
                 Vec::new()
             }
+            Err(_) => {
+                tracing::warn!("[{repo_id}] CVE scanning timed out after 10 minutes");
+                Vec::new()
+            }
         };
 
         // Stage 4: Pattern Scanning (GDPR + OAuth)
@@ -241,21 +247,6 @@ impl PipelineOrchestrator {
             Err(e) => tracing::warn!("[{repo_id}] Lint scanning failed: {e}"),
         }
 
-        // Stage 4c: LLM Code Review (only on incremental scans)
-        if let Some(old_sha) = &repo.last_scanned_commit {
-            tracing::info!("[{repo_id}] Stage 4c: LLM Code Review");
-            self.update_phase(scan_run_id, "code_review").await;
-            let review_output = async {
-                let reviewer = CodeReviewScanner::new(self.llm.clone());
-                reviewer
-                    .review_diff(&repo_path, &repo_id, old_sha, &current_sha)
-                    .await
-            }
-            .instrument(tracing::info_span!("stage_code_review"))
-            .await;
-            all_findings.extend(review_output.findings);
-        }
-
         // Stage 4.5: Graph Building
         tracing::info!("[{repo_id}] Stage 4.5: Graph Building");
         self.update_phase(scan_run_id, "graph_building").await;
@@ -331,20 +322,67 @@ impl PipelineOrchestrator {
                 .await?;
         }
 
-        // Persist CVE alerts (upsert by cve_id + repo_id)
-        for alert in &cve_alerts {
-            let filter = doc! {
-                "cve_id": &alert.cve_id,
-                "repo_id": &alert.repo_id,
-            };
-            let update = mongodb::bson::to_document(alert)
-                .map(|d| doc! { "$set": d })
-                .unwrap_or_else(|_| doc! {});
-            self.db
-                .cve_alerts()
-                .update_one(filter, update)
-                .upsert(true)
-                .await?;
+        // Persist CVE alerts and create notifications
+        {
+            use compliance_core::models::notification::{parse_severity, CveNotification};
+
+            let repo_name = repo.name.clone();
+            let mut new_notif_count = 0u32;
+
+            for alert in &cve_alerts {
+                // Upsert the alert
+                let filter = doc! {
+                    "cve_id": &alert.cve_id,
+                    "repo_id": &alert.repo_id,
+                };
+                let update = mongodb::bson::to_document(alert)
+                    .map(|d| doc! { "$set": d })
+                    .unwrap_or_else(|_| doc! {});
+                self.db
+                    .cve_alerts()
+                    .update_one(filter, update)
+                    .upsert(true)
+                    .await?;
+
+                // Create notification (dedup by cve_id + repo + package + version)
+                let notif_filter = doc! {
+                    "cve_id": &alert.cve_id,
+                    "repo_id": &alert.repo_id,
+                    "package_name": &alert.affected_package,
+                    "package_version": &alert.affected_version,
+                };
+                let severity = parse_severity(alert.severity.as_deref(), alert.cvss_score);
+                let mut notification = CveNotification::new(
+                    alert.cve_id.clone(),
+                    repo_id.clone(),
+                    repo_name.clone(),
+                    alert.affected_package.clone(),
+                    alert.affected_version.clone(),
+                    severity,
+                );
+                notification.cvss_score = alert.cvss_score;
+                notification.summary = alert.summary.clone();
+                notification.url = Some(format!("https://osv.dev/vulnerability/{}", alert.cve_id));
+
+                let notif_update = doc! {
+                    "$setOnInsert": mongodb::bson::to_bson(&notification).unwrap_or_default()
+                };
+                if let Ok(result) = self
+                    .db
+                    .cve_notifications()
+                    .update_one(notif_filter, notif_update)
+                    .upsert(true)
+                    .await
+                {
+                    if result.upserted_id.is_some() {
+                        new_notif_count += 1;
+                    }
+                }
+            }
+
+            if new_notif_count > 0 {
+                tracing::info!("[{repo_id}] Created {new_notif_count} CVE notification(s)");
+            }
         }
 
         // Stage 6: Issue Creation

compliance-agent/src/pipeline/patterns.rs

@@ -33,6 +33,7 @@ struct PatternRule {
     file_extensions: Vec<String>,
 }
 
+#[allow(clippy::new_without_default)]
 impl GdprPatternScanner {
     pub fn new() -> Self {
         let patterns = vec![
@@ -98,6 +99,7 @@ impl Scanner for GdprPatternScanner {
     }
 }
 
+#[allow(clippy::new_without_default)]
 impl OAuthPatternScanner {
     pub fn new() -> Self {
         let patterns = vec![

compliance-agent/src/pipeline/pr_review.rs

@@ -1,5 +1,6 @@
 use compliance_core::models::*;
 
+use super::dedup::compute_fingerprint;
 use super::orchestrator::PipelineOrchestrator;
 use crate::error::AgentError;
 use crate::pipeline::code_review::CodeReviewScanner;
@@ -89,12 +90,37 @@ impl PipelineOrchestrator {
             return Ok(());
         }
 
+        // Dedup findings by fingerprint to avoid duplicate comments
+        let mut seen_fps = std::collections::HashSet::new();
+        let mut unique_findings: Vec<&Finding> = Vec::new();
+        for finding in &pr_findings {
+            let fp = compute_fingerprint(&[
+                repo_id,
+                &pr_number.to_string(),
+                finding.file_path.as_deref().unwrap_or(""),
+                &finding.line_number.unwrap_or(0).to_string(),
+                &finding.title,
+            ]);
+            if seen_fps.insert(fp) {
+                unique_findings.push(finding);
+            }
+        }
+
+        let pr_findings = unique_findings;
+
         // Build review comments from findings
         let mut review_comments = Vec::new();
         for finding in &pr_findings {
             if let (Some(path), Some(line)) = (&finding.file_path, finding.line_number) {
+                let fp = compute_fingerprint(&[
+                    repo_id,
+                    &pr_number.to_string(),
+                    path,
+                    &line.to_string(),
+                    &finding.title,
+                ]);
                 let comment_body = format!(
-                    "**[{}] {}**\n\n{}\n\n*Scanner: {} | {}*",
+                    "**[{}] {}**\n\n{}\n\n*Scanner: {} | {}*\n\n<!-- compliance-fp:{fp} -->",
                     finding.severity,
                     finding.title,
                     finding.description,
@@ -123,6 +149,17 @@ impl PipelineOrchestrator {
                 .join("\n"),
         );
 
+        if review_comments.is_empty() {
+            // All findings were on files/lines we can't comment on inline
+            if let Err(e) = tracker
+                .create_pr_review(owner, tracker_repo_name, pr_number, &summary, Vec::new())
+                .await
+            {
+                tracing::warn!("[{repo_id}] Failed to post PR review summary: {e}");
+            }
+            return Ok(());
+        }
+
         if let Err(e) = tracker
             .create_pr_review(
                 owner,

compliance-agent/src/pipeline/sbom/syft.rs

@@ -5,20 +5,26 @@ use compliance_core::CoreError;
 
 #[tracing::instrument(skip_all, fields(repo_id = %repo_id))]
 pub(super) async fn run_syft(repo_path: &Path, repo_id: &str) -> Result<Vec<SbomEntry>, CoreError> {
-    let output = tokio::process::Command::new("syft")
-        .arg(repo_path)
-        .args(["-o", "cyclonedx-json"])
-        // Enable remote license lookups for all ecosystems
-        .env("SYFT_GOLANG_SEARCH_REMOTE_LICENSES", "true")
-        .env("SYFT_JAVASCRIPT_SEARCH_REMOTE_LICENSES", "true")
-        .env("SYFT_PYTHON_SEARCH_REMOTE_LICENSES", "true")
-        .env("SYFT_JAVA_USE_NETWORK", "true")
-        .output()
-        .await
-        .map_err(|e| CoreError::Scanner {
-            scanner: "syft".to_string(),
-            source: Box::new(e),
-        })?;
+    let output = tokio::time::timeout(
+        std::time::Duration::from_secs(300),
+        tokio::process::Command::new("syft")
+            .arg(repo_path)
+            .args(["-o", "cyclonedx-json"])
+            .env("SYFT_GOLANG_SEARCH_REMOTE_LICENSES", "true")
+            .env("SYFT_JAVASCRIPT_SEARCH_REMOTE_LICENSES", "true")
+            .env("SYFT_PYTHON_SEARCH_REMOTE_LICENSES", "true")
+            .env("SYFT_JAVA_USE_NETWORK", "true")
+            .output(),
+    )
+    .await
+    .map_err(|_| CoreError::Scanner {
+        scanner: "syft".to_string(),
+        source: "timed out after 5 minutes".into(),
+    })?
+    .map_err(|e| CoreError::Scanner {
+        scanner: "syft".to_string(),
+        source: Box::new(e),
+    })?;
 
     if !output.status.success() {
         let stderr = String::from_utf8_lossy(&output.stderr);

compliance-agent/src/pipeline/semgrep.rs

@@ -19,15 +19,30 @@ impl Scanner for SemgrepScanner {
 
     #[tracing::instrument(skip_all)]
     async fn scan(&self, repo_path: &Path, repo_id: &str) -> Result<ScanOutput, CoreError> {
-        let output = tokio::process::Command::new("semgrep")
-            .args(["--config=auto", "--json", "--quiet"])
-            .arg(repo_path)
-            .output()
-            .await
-            .map_err(|e| CoreError::Scanner {
-                scanner: "semgrep".to_string(),
-                source: Box::new(e),
-            })?;
+        let output = tokio::time::timeout(
+            std::time::Duration::from_secs(600),
+            tokio::process::Command::new("semgrep")
+                .args([
+                    "--config=auto",
+                    "--json",
+                    "--quiet",
+                    "--max-memory",
+                    "500",
+                    "--jobs",
+                    "1",
+                ])
+                .arg(repo_path)
+                .output(),
+        )
+        .await
+        .map_err(|_| CoreError::Scanner {
+            scanner: "semgrep".to_string(),
+            source: "timed out after 10 minutes".into(),
+        })?
+        .map_err(|e| CoreError::Scanner {
+            scanner: "semgrep".to_string(),
+            source: Box::new(e),
+        })?;
 
         if !output.status.success() && output.stdout.is_empty() {
             let stderr = String::from_utf8_lossy(&output.stderr);

compliance-agent/src/rag/pipeline.rs

@@ -6,11 +6,16 @@ use compliance_core::models::embedding::{CodeEmbedding, EmbeddingBuildRun, Embed
 use compliance_core::models::graph::CodeNode;
 use compliance_graph::graph::chunking::extract_chunks;
 use compliance_graph::graph::embedding_store::EmbeddingStore;
+use futures_util::stream::{FuturesUnordered, StreamExt};
 use tracing::{error, info};
 
 use crate::error::AgentError;
 use crate::llm::LlmClient;
 
+const EMBED_BATCH_SIZE: usize = 20;
+const EMBED_CONCURRENCY: usize = 4;
+const EMBED_FLUSH_EVERY: usize = 200;
+
 /// RAG pipeline for building embeddings and performing retrieval
 pub struct RagPipeline {
     llm: Arc<LlmClient>,
@@ -77,25 +82,33 @@ impl RagPipeline {
             .await
             .map_err(|e| AgentError::Other(format!("Failed to delete old embeddings: {e}")))?;
 
-        // Step 3: Batch embed (small batches to stay within model limits)
-        let batch_size = 20;
-        let mut all_embeddings = Vec::new();
+        // Step 3: Batch embed with bounded concurrency. Flush to Mongo and
+        // update progress periodically so the dashboard can show live status.
+        let mut pending = Vec::with_capacity(EMBED_FLUSH_EVERY);
         let mut embedded_count = 0u32;
 
-        for batch_start in (0..chunks.len()).step_by(batch_size) {
-            let batch_end = (batch_start + batch_size).min(chunks.len());
-            let batch_chunks = &chunks[batch_start..batch_end];
+        // Build the list of batch indices to process.
+        let batches: Vec<(usize, usize)> = (0..chunks.len())
+            .step_by(EMBED_BATCH_SIZE)
+            .map(|start| (start, (start + EMBED_BATCH_SIZE).min(chunks.len())))
+            .collect();
 
-            // Prepare texts: context_header + content
-            let texts: Vec<String> = batch_chunks
-                .iter()
-                .map(|c| format!("{}\n{}", c.context_header, c.content))
-                .collect();
+        let mut batch_iter = batches.into_iter();
+        let mut in_flight = FuturesUnordered::new();
 
-            match self.llm.embed(texts).await {
-                Ok(vectors) => {
+        // Prime up to EMBED_CONCURRENCY batches.
+        for _ in 0..EMBED_CONCURRENCY {
+            if let Some((start, end)) = batch_iter.next() {
+                in_flight.push(self.embed_batch(&chunks[start..end], start, end));
+            }
+        }
+
+        while let Some(result) = in_flight.next().await {
+            match result {
+                Ok((start, end, vectors)) => {
+                    let batch_chunks = &chunks[start..end];
                     for (chunk, embedding) in batch_chunks.iter().zip(vectors) {
-                        all_embeddings.push(CodeEmbedding {
+                        pending.push(CodeEmbedding {
                             id: None,
                             repo_id: repo_id.to_string(),
                             graph_build_id: graph_build_id.to_string(),
@@ -113,9 +126,45 @@ impl RagPipeline {
                         });
                     }
                     embedded_count += batch_chunks.len() as u32;
+
+                    // Flush pending embeddings to Mongo periodically and update progress.
+                    if pending.len() >= EMBED_FLUSH_EVERY {
+                        self.embedding_store
+                            .store_embeddings(&pending)
+                            .await
+                            .map_err(|e| {
+                                AgentError::Other(format!("Failed to store embeddings: {e}"))
+                            })?;
+                        pending.clear();
+                    }
+
+                    // Always update the progress counter on the build doc — even if
+                    // we haven't flushed embeddings yet — so the UI shows movement.
+                    if let Err(e) = self
+                        .embedding_store
+                        .update_build(
+                            repo_id,
+                            graph_build_id,
+                            EmbeddingBuildStatus::Running,
+                            embedded_count,
+                            None,
+                        )
+                        .await
+                    {
+                        error!("[{repo_id}] Failed to update build progress: {e}");
+                    }
+
+                    // Queue the next batch to keep concurrency saturated.
+                    if let Some((s, e)) = batch_iter.next() {
+                        in_flight.push(self.embed_batch(&chunks[s..e], s, e));
+                    }
                 }
                 Err(e) => {
                     error!("[{repo_id}] Embedding batch failed: {e}");
+                    // Flush whatever we have so partial progress isn't lost.
+                    if !pending.is_empty() {
+                        let _ = self.embedding_store.store_embeddings(&pending).await;
+                    }
                     build.status = EmbeddingBuildStatus::Failed;
                     build.error_message = Some(e.to_string());
                     build.completed_at = Some(Utc::now());
@@ -134,11 +183,13 @@ impl RagPipeline {
             }
         }
 
-        // Step 4: Store all embeddings
-        self.embedding_store
-            .store_embeddings(&all_embeddings)
-            .await
-            .map_err(|e| AgentError::Other(format!("Failed to store embeddings: {e}")))?;
+        // Step 4: Flush any remaining embeddings
+        if !pending.is_empty() {
+            self.embedding_store
+                .store_embeddings(&pending)
+                .await
+                .map_err(|e| AgentError::Other(format!("Failed to store embeddings: {e}")))?;
+        }
 
         // Step 5: Update build status
         build.status = EmbeddingBuildStatus::Completed;
@@ -161,4 +212,21 @@ impl RagPipeline {
         );
         Ok(build)
     }
+
+    /// Embed one batch of chunks. Returns the (start, end, vectors) tuple so
+    /// out-of-order completion from `FuturesUnordered` can still be reconciled
+    /// against the original chunk slice.
+    async fn embed_batch(
+        &self,
+        batch_chunks: &[compliance_graph::graph::chunking::CodeChunk],
+        start: usize,
+        end: usize,
+    ) -> Result<(usize, usize, Vec<Vec<f64>>), AgentError> {
+        let texts: Vec<String> = batch_chunks
+            .iter()
+            .map(|c| format!("{}\n{}", c.context_header, c.content))
+            .collect();
+        let vectors = self.llm.embed(texts).await?;
+        Ok((start, end, vectors))
+    }
 }

compliance-agent/src/scheduler.rs

@@ -4,8 +4,14 @@ use tokio_cron_scheduler::{Job, JobScheduler};
 use compliance_core::models::ScanTrigger;
 
 use crate::agent::ComplianceAgent;
+use crate::database::Database;
 use crate::error::AgentError;
 
+/// Default tenant the scheduler runs against when `SCHEDULER_TENANT_IDS`
+/// isn't set. Matches the dev-injector default so a bare `cargo run` has
+/// the scheduler scanning whatever lives in `<prefix>_dev`.
+const DEFAULT_SCHEDULER_TENANT_ID: &str = "dev";
+
 pub async fn start_scheduler(agent: &ComplianceAgent) -> Result<(), AgentError> {
     let sched = JobScheduler::new()
         .await
@@ -18,7 +24,9 @@ pub async fn start_scheduler(agent: &ComplianceAgent) -> Result<(), AgentError>
         let agent = scan_agent.clone();
         Box::pin(async move {
             tracing::info!("Scheduled scan triggered");
-            scan_all_repos(&agent).await;
+            for tenant_id in scheduler_tenants() {
+                scan_all_repos(&agent, &tenant_id).await;
+            }
         })
     })
     .map_err(|e| AgentError::Scheduler(format!("Failed to create scan job: {e}")))?;
@@ -34,7 +42,9 @@ pub async fn start_scheduler(agent: &ComplianceAgent) -> Result<(), AgentError>
         let agent = cve_agent.clone();
         Box::pin(async move {
             tracing::info!("CVE monitor triggered");
-            monitor_cves(&agent).await;
+            for tenant_id in scheduler_tenants() {
+                monitor_cves(&agent, &tenant_id).await;
+            }
         })
     })
     .map_err(|e| AgentError::Scheduler(format!("Failed to create CVE monitor job: {e}")))?;
@@ -48,8 +58,9 @@ pub async fn start_scheduler(agent: &ComplianceAgent) -> Result<(), AgentError>
         .await
         .map_err(|e| AgentError::Scheduler(format!("Failed to start scheduler: {e}")))?;
 
+    let tenants = scheduler_tenants();
     tracing::info!(
-        "Scheduler started: scans='{}', CVE monitor='{}'",
+        "Scheduler started: scans='{}', CVE monitor='{}', tenants={tenants:?}",
         agent.config.scan_schedule,
         agent.config.cve_monitor_schedule,
     );
@@ -60,13 +71,47 @@ pub async fn start_scheduler(agent: &ComplianceAgent) -> Result<(), AgentError>
     }
 }
 
-async fn scan_all_repos(agent: &ComplianceAgent) {
+/// Tenants the scheduler iterates each tick. From `SCHEDULER_TENANT_IDS`
+/// (comma-separated), or `DEFAULT_SCHEDULER_TENANT_ID` if unset. M7.2-D
+/// will replace this with a pull from the tenant-registry.
+fn scheduler_tenants() -> Vec<String> {
+    std::env::var("SCHEDULER_TENANT_IDS")
+        .ok()
+        .map(|s| {
+            s.split(',')
+                .map(str::trim)
+                .filter(|s| !s.is_empty())
+                .map(String::from)
+                .collect::<Vec<_>>()
+        })
+        .filter(|v| !v.is_empty())
+        .unwrap_or_else(|| vec![DEFAULT_SCHEDULER_TENANT_ID.to_string()])
+}
+
+/// Resolve the per-tenant database. Logs and returns `None` on failure
+/// so the loop in the caller can continue with other tenants.
+async fn tenant_db(agent: &ComplianceAgent, tenant_id: &str) -> Option<Database> {
+    match agent.db_pool.for_tenant_id(tenant_id).await {
+        Ok(db) => Some(db),
+        Err(e) => {
+            tracing::error!("Scheduler: cannot open tenant database '{tenant_id}': {e}");
+            None
+        }
+    }
+}
+
+async fn scan_all_repos(agent: &ComplianceAgent, tenant_id: &str) {
     use futures_util::StreamExt;
 
-    let cursor = match agent.db.repositories().find(doc! {}).await {
+    let db = match tenant_db(agent, tenant_id).await {
+        Some(db) => db,
+        None => return,
+    };
+
+    let cursor = match db.repositories().find(doc! {}).await {
         Ok(c) => c,
         Err(e) => {
-            tracing::error!("Failed to list repos for scheduled scan: {e}");
+            tracing::error!("Failed to list repos for tenant '{tenant_id}': {e}");
             return;
         }
     };
@@ -75,31 +120,175 @@ async fn scan_all_repos(agent: &ComplianceAgent) {
 
     for repo in repos {
         let repo_id = repo.id.map(|id| id.to_hex()).unwrap_or_default();
-        if let Err(e) = agent.run_scan(&repo_id, ScanTrigger::Scheduled).await {
-            tracing::error!("Scheduled scan failed for {}: {e}", repo.name);
+        if let Err(e) = agent
+            .run_scan(tenant_id, &repo_id, ScanTrigger::Scheduled)
+            .await
+        {
+            tracing::error!(
+                "Scheduled scan failed for {} (tenant '{tenant_id}'): {e}",
+                repo.name
+            );
         }
     }
 }
 
-async fn monitor_cves(agent: &ComplianceAgent) {
+async fn monitor_cves(agent: &ComplianceAgent, tenant_id: &str) {
+    use compliance_core::models::notification::{parse_severity, CveNotification};
+    use compliance_core::models::SbomEntry;
     use futures_util::StreamExt;
 
-    // Re-scan all SBOM entries for new CVEs
-    let cursor = match agent.db.sbom_entries().find(doc! {}).await {
+    let db = match tenant_db(agent, tenant_id).await {
+        Some(db) => db,
+        None => return,
+    };
+
+    // Fetch all SBOM entries grouped by repo
+    let cursor = match db.sbom_entries().find(doc! {}).await {
         Ok(c) => c,
         Err(e) => {
-            tracing::error!("Failed to list SBOM entries for CVE monitoring: {e}");
+            tracing::error!("CVE monitor: failed to list SBOM entries for '{tenant_id}': {e}");
             return;
         }
     };
-
-    let entries: Vec<_> = cursor.filter_map(|r| async { r.ok() }).collect().await;
-
+    let entries: Vec<SbomEntry> = cursor.filter_map(|r| async { r.ok() }).collect().await;
     if entries.is_empty() {
+        tracing::debug!("CVE monitor: no SBOM entries for tenant '{tenant_id}', skipping");
         return;
     }
 
-    tracing::info!("CVE monitor: checking {} dependencies", entries.len());
-    // The actual CVE checking is handled by the CveScanner in the pipeline
-    // This is a simplified version that just logs the activity
+    tracing::info!(
+        "CVE monitor: checking {} dependencies for new CVEs (tenant '{tenant_id}')",
+        entries.len()
+    );
+
+    // Build a repo_id → repo_name lookup
+    let repo_ids: std::collections::HashSet<String> =
+        entries.iter().map(|e| e.repo_id.clone()).collect();
+    let mut repo_names: std::collections::HashMap<String, String> =
+        std::collections::HashMap::new();
+    for rid in &repo_ids {
+        if let Ok(oid) = mongodb::bson::oid::ObjectId::parse_str(rid) {
+            if let Ok(Some(repo)) = db.repositories().find_one(doc! { "_id": oid }).await {
+                repo_names.insert(rid.clone(), repo.name.clone());
+            }
+        }
+    }
+
+    // Use the existing CveScanner to query OSV.dev
+    let nvd_key = agent.config.nvd_api_key.as_ref().map(|k| {
+        use secrecy::ExposeSecret;
+        k.expose_secret().to_string()
+    });
+    let scanner = crate::pipeline::cve::CveScanner::new(
+        agent.http.clone(),
+        agent.config.searxng_url.clone(),
+        nvd_key,
+    );
+
+    // Group entries by repo for scanning
+    let mut entries_by_repo: std::collections::HashMap<String, Vec<SbomEntry>> =
+        std::collections::HashMap::new();
+    for entry in entries {
+        entries_by_repo
+            .entry(entry.repo_id.clone())
+            .or_default()
+            .push(entry);
+    }
+
+    let mut new_notifications = 0u32;
+
+    for (repo_id, mut repo_entries) in entries_by_repo {
+        let repo_name = repo_names
+            .get(&repo_id)
+            .cloned()
+            .unwrap_or_else(|| repo_id.clone());
+
+        // Scan dependencies for CVEs
+        let alerts = match scanner.scan_dependencies(&repo_id, &mut repo_entries).await {
+            Ok(a) => a,
+            Err(e) => {
+                tracing::warn!("CVE monitor: scan failed for {repo_name}: {e}");
+                continue;
+            }
+        };
+
+        // Upsert CVE alerts (existing logic)
+        for alert in &alerts {
+            let filter = doc! { "cve_id": &alert.cve_id, "repo_id": &alert.repo_id };
+            let update = doc! { "$setOnInsert": mongodb::bson::to_bson(alert).unwrap_or_default() };
+            let _ = db
+                .cve_alerts()
+                .update_one(filter, update)
+                .upsert(true)
+                .await;
+        }
+
+        // Update SBOM entries with discovered vulnerabilities
+        for entry in &repo_entries {
+            if entry.known_vulnerabilities.is_empty() {
+                continue;
+            }
+            if let Some(entry_id) = &entry.id {
+                let _ = db
+                    .sbom_entries()
+                    .update_one(
+                        doc! { "_id": entry_id },
+                        doc! { "$set": {
+                            "known_vulnerabilities": mongodb::bson::to_bson(&entry.known_vulnerabilities).unwrap_or_default(),
+                            "updated_at": mongodb::bson::DateTime::now(),
+                        }},
+                    )
+                    .await;
+            }
+        }
+
+        // Create notifications for NEW CVEs (dedup against existing notifications)
+        for alert in &alerts {
+            let filter = doc! {
+                "cve_id": &alert.cve_id,
+                "repo_id": &alert.repo_id,
+                "package_name": &alert.affected_package,
+                "package_version": &alert.affected_version,
+            };
+            // Only insert if not already exists (upsert with $setOnInsert)
+            let severity = parse_severity(alert.severity.as_deref(), alert.cvss_score);
+            let mut notification = CveNotification::new(
+                alert.cve_id.clone(),
+                repo_id.clone(),
+                repo_name.clone(),
+                alert.affected_package.clone(),
+                alert.affected_version.clone(),
+                severity,
+            );
+            notification.cvss_score = alert.cvss_score;
+            notification.summary = alert.summary.clone();
+            notification.url = Some(format!("https://osv.dev/vulnerability/{}", alert.cve_id));
+
+            let update = doc! {
+                "$setOnInsert": mongodb::bson::to_bson(&notification).unwrap_or_default()
+            };
+            match db
+                .cve_notifications()
+                .update_one(filter, update)
+                .upsert(true)
+                .await
+            {
+                Ok(result) if result.upserted_id.is_some() => {
+                    new_notifications += 1;
+                }
+                Err(e) => {
+                    tracing::warn!("CVE monitor: failed to create notification: {e}");
+                }
+                _ => {} // Already exists
+            }
+        }
+    }
+
+    if new_notifications > 0 {
+        tracing::info!(
+            "CVE monitor: created {new_notifications} new notification(s) for tenant '{tenant_id}'"
+        );
+    } else {
+        tracing::info!("CVE monitor: no new CVEs found for tenant '{tenant_id}'");
+    }
 }

compliance-agent/src/trackers/gitea.rs

@@ -98,7 +98,8 @@ impl IssueTracker for GiteaTracker {
             _ => "open",
         };
 
-        self.http
+        let resp = self
+            .http
             .patch(&url)
             .header(
                 "Authorization",
@@ -109,6 +110,14 @@ impl IssueTracker for GiteaTracker {
             .await
             .map_err(|e| CoreError::IssueTracker(format!("Gitea update issue failed: {e}")))?;
 
+        if !resp.status().is_success() {
+            let status = resp.status();
+            let text = resp.text().await.unwrap_or_default();
+            return Err(CoreError::IssueTracker(format!(
+                "Gitea update issue returned {status}: {text}"
+            )));
+        }
+
         Ok(())
     }
 
@@ -123,7 +132,8 @@ impl IssueTracker for GiteaTracker {
             "/repos/{owner}/{repo}/issues/{external_id}/comments"
         ));
 
-        self.http
+        let resp = self
+            .http
             .post(&url)
             .header(
                 "Authorization",
@@ -134,6 +144,14 @@ impl IssueTracker for GiteaTracker {
             .await
             .map_err(|e| CoreError::IssueTracker(format!("Gitea add comment failed: {e}")))?;
 
+        if !resp.status().is_success() {
+            let status = resp.status();
+            let text = resp.text().await.unwrap_or_default();
+            return Err(CoreError::IssueTracker(format!(
+                "Gitea add comment returned {status}: {text}"
+            )));
+        }
+
         Ok(())
     }
 
@@ -158,7 +176,8 @@ impl IssueTracker for GiteaTracker {
             })
             .collect();
 
-        self.http
+        let resp = self
+            .http
             .post(&url)
             .header(
                 "Authorization",
@@ -173,6 +192,48 @@ impl IssueTracker for GiteaTracker {
             .await
             .map_err(|e| CoreError::IssueTracker(format!("Gitea PR review failed: {e}")))?;
 
+        if !resp.status().is_success() {
+            let status = resp.status();
+            let text = resp.text().await.unwrap_or_default();
+
+            // If inline comments caused the failure, retry with just the summary body
+            if !comments.is_empty() {
+                tracing::warn!(
+                    "Gitea PR review with inline comments failed ({status}): {text}, retrying as plain comment"
+                );
+                let fallback_url = self.api_url(&format!(
+                    "/repos/{owner}/{repo}/issues/{pr_number}/comments"
+                ));
+                let fallback_resp = self
+                    .http
+                    .post(&fallback_url)
+                    .header(
+                        "Authorization",
+                        format!("token {}", self.token.expose_secret()),
+                    )
+                    .json(&serde_json::json!({ "body": body }))
+                    .send()
+                    .await
+                    .map_err(|e| {
+                        CoreError::IssueTracker(format!("Gitea PR comment fallback failed: {e}"))
+                    })?;
+
+                if !fallback_resp.status().is_success() {
+                    let fb_status = fallback_resp.status();
+                    let fb_text = fallback_resp.text().await.unwrap_or_default();
+                    return Err(CoreError::IssueTracker(format!(
+                        "Gitea PR comment fallback returned {fb_status}: {fb_text}"
+                    )));
+                }
+
+                return Ok(());
+            }
+
+            return Err(CoreError::IssueTracker(format!(
+                "Gitea PR review returned {status}: {text}"
+            )));
+        }
+
         Ok(())
     }
 

compliance-agent/src/webhooks/gitea.rs

@@ -14,24 +14,30 @@ type HmacSha256 = Hmac<Sha256>;
 
 pub async fn handle_gitea_webhook(
     Extension(agent): Extension<Arc<ComplianceAgent>>,
-    Path(repo_id): Path<String>,
+    Path((tenant_id, repo_id)): Path<(String, String)>,
     headers: HeaderMap,
     body: Bytes,
 ) -> StatusCode {
-    // Look up the repo to get its webhook secret
+    // Look up the repo in the tenant's database to get its webhook secret
     let oid = match mongodb::bson::oid::ObjectId::parse_str(&repo_id) {
         Ok(oid) => oid,
         Err(_) => return StatusCode::NOT_FOUND,
     };
-    let repo = match agent
-        .db
+    let db = match agent.db_pool.for_tenant_id(&tenant_id).await {
+        Ok(db) => db,
+        Err(e) => {
+            tracing::warn!("Gitea webhook: cannot open tenant database '{tenant_id}': {e}");
+            return StatusCode::NOT_FOUND;
+        }
+    };
+    let repo = match db
         .repositories()
         .find_one(mongodb::bson::doc! { "_id": oid })
         .await
     {
         Ok(Some(repo)) => repo,
         _ => {
-            tracing::warn!("Gitea webhook: repo {repo_id} not found");
+            tracing::warn!("Gitea webhook: repo {repo_id} not found in tenant '{tenant_id}'");
             return StatusCode::NOT_FOUND;
         }
     };
@@ -66,15 +72,21 @@ pub async fn handle_gitea_webhook(
         "push" => {
             let agent_clone = (*agent).clone();
             let repo_id = repo_id.clone();
+            let tenant_id = tenant_id.clone();
             tokio::spawn(async move {
-                tracing::info!("Gitea push webhook: triggering scan for {repo_id}");
-                if let Err(e) = agent_clone.run_scan(&repo_id, ScanTrigger::Webhook).await {
+                tracing::info!(
+                    "Gitea push webhook: triggering scan for {repo_id} in tenant {tenant_id}"
+                );
+                if let Err(e) = agent_clone
+                    .run_scan(&tenant_id, &repo_id, ScanTrigger::Webhook)
+                    .await
+                {
                     tracing::error!("Webhook-triggered scan failed: {e}");
                 }
             });
             StatusCode::OK
         }
-        "pull_request" => handle_pull_request(agent, &repo_id, &payload).await,
+        "pull_request" => handle_pull_request(agent, &tenant_id, &repo_id, &payload).await,
         _ => {
             tracing::debug!("Gitea webhook: ignoring event '{event}'");
             StatusCode::OK
@@ -84,6 +96,7 @@ pub async fn handle_gitea_webhook(
 
 async fn handle_pull_request(
     agent: Arc<ComplianceAgent>,
+    tenant_id: &str,
     repo_id: &str,
     payload: &serde_json::Value,
 ) -> StatusCode {
@@ -106,13 +119,14 @@ async fn handle_pull_request(
     }
 
     let repo_id = repo_id.to_string();
+    let tenant_id = tenant_id.to_string();
     let head_sha = head_sha.to_string();
     let base_sha = base_sha.to_string();
     let agent_clone = (*agent).clone();
     tokio::spawn(async move {
         tracing::info!("Gitea PR webhook: reviewing PR #{pr_number} on {repo_id}");
         if let Err(e) = agent_clone
-            .run_pr_review(&repo_id, pr_number, &base_sha, &head_sha)
+            .run_pr_review(&tenant_id, &repo_id, pr_number, &base_sha, &head_sha)
             .await
         {
             tracing::error!("PR review failed for #{pr_number}: {e}");

compliance-agent/src/webhooks/github.rs

@@ -14,24 +14,30 @@ type HmacSha256 = Hmac<Sha256>;
 
 pub async fn handle_github_webhook(
     Extension(agent): Extension<Arc<ComplianceAgent>>,
-    Path(repo_id): Path<String>,
+    Path((tenant_id, repo_id)): Path<(String, String)>,
     headers: HeaderMap,
     body: Bytes,
 ) -> StatusCode {
-    // Look up the repo to get its webhook secret
+    // Look up the repo in the tenant's database to get its webhook secret
     let oid = match mongodb::bson::oid::ObjectId::parse_str(&repo_id) {
         Ok(oid) => oid,
         Err(_) => return StatusCode::NOT_FOUND,
     };
-    let repo = match agent
-        .db
+    let db = match agent.db_pool.for_tenant_id(&tenant_id).await {
+        Ok(db) => db,
+        Err(e) => {
+            tracing::warn!("GitHub webhook: cannot open tenant database '{tenant_id}': {e}");
+            return StatusCode::NOT_FOUND;
+        }
+    };
+    let repo = match db
         .repositories()
         .find_one(mongodb::bson::doc! { "_id": oid })
         .await
     {
         Ok(Some(repo)) => repo,
         _ => {
-            tracing::warn!("GitHub webhook: repo {repo_id} not found");
+            tracing::warn!("GitHub webhook: repo {repo_id} not found in tenant '{tenant_id}'");
             return StatusCode::NOT_FOUND;
         }
     };
@@ -66,15 +72,21 @@ pub async fn handle_github_webhook(
         "push" => {
             let agent_clone = (*agent).clone();
             let repo_id = repo_id.clone();
+            let tenant_id = tenant_id.clone();
             tokio::spawn(async move {
-                tracing::info!("GitHub push webhook: triggering scan for {repo_id}");
-                if let Err(e) = agent_clone.run_scan(&repo_id, ScanTrigger::Webhook).await {
+                tracing::info!(
+                    "GitHub push webhook: triggering scan for {repo_id} in tenant {tenant_id}"
+                );
+                if let Err(e) = agent_clone
+                    .run_scan(&tenant_id, &repo_id, ScanTrigger::Webhook)
+                    .await
+                {
                     tracing::error!("Webhook-triggered scan failed: {e}");
                 }
             });
             StatusCode::OK
         }
-        "pull_request" => handle_pull_request(agent, &repo_id, &payload).await,
+        "pull_request" => handle_pull_request(agent, &tenant_id, &repo_id, &payload).await,
         _ => {
             tracing::debug!("GitHub webhook: ignoring event '{event}'");
             StatusCode::OK
@@ -84,6 +96,7 @@ pub async fn handle_github_webhook(
 
 async fn handle_pull_request(
     agent: Arc<ComplianceAgent>,
+    tenant_id: &str,
     repo_id: &str,
     payload: &serde_json::Value,
 ) -> StatusCode {
@@ -105,13 +118,14 @@ async fn handle_pull_request(
     }
 
     let repo_id = repo_id.to_string();
+    let tenant_id = tenant_id.to_string();
     let head_sha = head_sha.to_string();
     let base_sha = base_sha.to_string();
     let agent_clone = (*agent).clone();
     tokio::spawn(async move {
         tracing::info!("GitHub PR webhook: reviewing PR #{pr_number} on {repo_id}");
         if let Err(e) = agent_clone
-            .run_pr_review(&repo_id, pr_number, &base_sha, &head_sha)
+            .run_pr_review(&tenant_id, &repo_id, pr_number, &base_sha, &head_sha)
             .await
         {
             tracing::error!("PR review failed for #{pr_number}: {e}");

compliance-agent/src/webhooks/gitlab.rs

@@ -10,24 +10,30 @@ use crate::agent::ComplianceAgent;
 
 pub async fn handle_gitlab_webhook(
     Extension(agent): Extension<Arc<ComplianceAgent>>,
-    Path(repo_id): Path<String>,
+    Path((tenant_id, repo_id)): Path<(String, String)>,
     headers: HeaderMap,
     body: Bytes,
 ) -> StatusCode {
-    // Look up the repo to get its webhook secret
+    // Look up the repo in the tenant's database to get its webhook secret
     let oid = match mongodb::bson::oid::ObjectId::parse_str(&repo_id) {
         Ok(oid) => oid,
         Err(_) => return StatusCode::NOT_FOUND,
     };
-    let repo = match agent
-        .db
+    let db = match agent.db_pool.for_tenant_id(&tenant_id).await {
+        Ok(db) => db,
+        Err(e) => {
+            tracing::warn!("GitLab webhook: cannot open tenant database '{tenant_id}': {e}");
+            return StatusCode::NOT_FOUND;
+        }
+    };
+    let repo = match db
         .repositories()
         .find_one(mongodb::bson::doc! { "_id": oid })
         .await
     {
         Ok(Some(repo)) => repo,
         _ => {
-            tracing::warn!("GitLab webhook: repo {repo_id} not found");
+            tracing::warn!("GitLab webhook: repo {repo_id} not found in tenant '{tenant_id}'");
             return StatusCode::NOT_FOUND;
         }
     };
@@ -59,15 +65,21 @@ pub async fn handle_gitlab_webhook(
         "push" => {
             let agent_clone = (*agent).clone();
             let repo_id = repo_id.clone();
+            let tenant_id = tenant_id.clone();
             tokio::spawn(async move {
-                tracing::info!("GitLab push webhook: triggering scan for {repo_id}");
-                if let Err(e) = agent_clone.run_scan(&repo_id, ScanTrigger::Webhook).await {
+                tracing::info!(
+                    "GitLab push webhook: triggering scan for {repo_id} in tenant {tenant_id}"
+                );
+                if let Err(e) = agent_clone
+                    .run_scan(&tenant_id, &repo_id, ScanTrigger::Webhook)
+                    .await
+                {
                     tracing::error!("Webhook-triggered scan failed: {e}");
                 }
             });
             StatusCode::OK
         }
-        "merge_request" => handle_merge_request(agent, &repo_id, &payload).await,
+        "merge_request" => handle_merge_request(agent, &tenant_id, &repo_id, &payload).await,
         _ => {
             tracing::debug!("GitLab webhook: ignoring event '{event_type}'");
             StatusCode::OK
@@ -77,6 +89,7 @@ pub async fn handle_gitlab_webhook(
 
 async fn handle_merge_request(
     agent: Arc<ComplianceAgent>,
+    tenant_id: &str,
     repo_id: &str,
     payload: &serde_json::Value,
 ) -> StatusCode {
@@ -101,13 +114,14 @@ async fn handle_merge_request(
     }
 
     let repo_id = repo_id.to_string();
+    let tenant_id = tenant_id.to_string();
     let head_sha = head_sha.to_string();
     let base_sha = base_sha.to_string();
     let agent_clone = (*agent).clone();
     tokio::spawn(async move {
         tracing::info!("GitLab MR webhook: reviewing MR !{mr_iid} on {repo_id}");
         if let Err(e) = agent_clone
-            .run_pr_review(&repo_id, mr_iid, &base_sha, &head_sha)
+            .run_pr_review(&tenant_id, &repo_id, mr_iid, &base_sha, &head_sha)
             .await
         {
             tracing::error!("MR review failed for !{mr_iid}: {e}");

compliance-agent/src/webhooks/server.rs

@@ -9,17 +9,21 @@ use crate::webhooks::{gitea, github, gitlab};
 
 pub async fn start_webhook_server(agent: &ComplianceAgent) -> Result<(), AgentError> {
     let app = Router::new()
-        // Per-repo webhook URLs: /webhook/{platform}/{repo_id}
+        // Per-tenant per-repo webhook URLs: /webhook/{tenant_id}/{platform}/{repo_id}
+        // The tenant_id is resolved from the URL path because webhooks
+        // arrive without a JWT — they're authenticated via per-repo HMAC,
+        // not via the tenant gate. The dashboard surfaces the full URL
+        // including the tenant_id when the repo is registered.
         .route(
-            "/webhook/github/{repo_id}",
+            "/webhook/{tenant_id}/github/{repo_id}",
             post(github::handle_github_webhook),
         )
         .route(
-            "/webhook/gitlab/{repo_id}",
+            "/webhook/{tenant_id}/gitlab/{repo_id}",
             post(gitlab::handle_gitlab_webhook),
         )
         .route(
-            "/webhook/gitea/{repo_id}",
+            "/webhook/{tenant_id}/gitea/{repo_id}",
             post(gitea::handle_gitea_webhook),
         )
         .layer(Extension(Arc::new(agent.clone())));

compliance-agent/tests/common/mod.rs

@@ -1,3 +1,177 @@
-// Shared test helpers for compliance-agent integration tests.
+// Shared test harness for E2E / integration tests.
 //
-// Add database mocks, fixtures, and test utilities here.
+// Spins up the agent API server on a random port with an isolated test
+// database. Each test gets a fresh database that is dropped on cleanup.
+
+use std::sync::Arc;
+
+use compliance_agent::agent::ComplianceAgent;
+use compliance_agent::api;
+use compliance_agent::database::DatabasePool;
+use compliance_core::AgentConfig;
+use secrecy::SecretString;
+
+/// A running test server with a unique database.
+pub struct TestServer {
+    pub base_url: String,
+    pub client: reqwest::Client,
+    db_name: String,
+    mongodb_uri: String,
+}
+
+impl TestServer {
+    /// Start an agent API server on a random port with an isolated database.
+    pub async fn start() -> Self {
+        let mongodb_uri = std::env::var("TEST_MONGODB_URI")
+            .unwrap_or_else(|_| "mongodb://root:example@localhost:27017/?authSource=admin".into());
+
+        // Unique database name per test run to avoid collisions
+        let db_name = format!("test_{}", uuid::Uuid::new_v4().simple());
+
+        let db_pool = DatabasePool::connect(&mongodb_uri, &db_name)
+            .await
+            .expect("Failed to build DatabasePool");
+
+        let config = AgentConfig {
+            mongodb_uri: mongodb_uri.clone(),
+            mongodb_database: db_name.clone(),
+            litellm_url: std::env::var("TEST_LITELLM_URL")
+                .unwrap_or_else(|_| "http://localhost:4000".into()),
+            litellm_api_key: SecretString::from(String::new()),
+            litellm_model: "gpt-4o".into(),
+            litellm_embed_model: "text-embedding-3-small".into(),
+            agent_port: 0, // not used — we bind ourselves
+            scan_schedule: String::new(),
+            cve_monitor_schedule: String::new(),
+            git_clone_base_path: "/tmp/compliance-scanner-tests/repos".into(),
+            ssh_key_path: "/tmp/compliance-scanner-tests/ssh/id_ed25519".into(),
+            github_token: None,
+            github_webhook_secret: None,
+            gitlab_url: None,
+            gitlab_token: None,
+            gitlab_webhook_secret: None,
+            jira_url: None,
+            jira_email: None,
+            jira_api_token: None,
+            jira_project_key: None,
+            searxng_url: None,
+            nvd_api_key: None,
+            keycloak_url: None,
+            keycloak_realm: None,
+            keycloak_admin_username: None,
+            keycloak_admin_password: None,
+            pentest_verification_email: None,
+            pentest_imap_host: None,
+            pentest_imap_port: None,
+            pentest_imap_tls: false,
+            pentest_imap_username: None,
+            pentest_imap_password: None,
+        };
+
+        let agent = ComplianceAgent::new(config, db_pool);
+
+        // Build the router with the agent extension. After M7.2-B every
+        // handler takes a TenantCtx extractor; without KC in the test
+        // harness, the dev-tenant injector mounts a synthetic context so
+        // tests run end-to-end against `<db_name>_dev`.
+        let app = api::routes::build_router()
+            .layer(axum::extract::Extension(Arc::new(agent)))
+            .layer(axum::middleware::from_fn(api::server::inject_dev_tenant))
+            .layer(tower_http::cors::CorsLayer::permissive());
+
+        // Bind to port 0 to get a random available port
+        let listener = tokio::net::TcpListener::bind("127.0.0.1:0")
+            .await
+            .expect("Failed to bind test server");
+        let port = listener.local_addr().expect("no local addr").port();
+
+        tokio::spawn(async move {
+            axum::serve(listener, app).await.ok();
+        });
+
+        let base_url = format!("http://127.0.0.1:{port}");
+        let client = reqwest::Client::builder()
+            .timeout(std::time::Duration::from_secs(30))
+            .build()
+            .expect("Failed to build HTTP client");
+
+        // Wait for server to be ready
+        for _ in 0..50 {
+            if client
+                .get(format!("{base_url}/api/v1/health"))
+                .send()
+                .await
+                .is_ok()
+            {
+                break;
+            }
+            tokio::time::sleep(std::time::Duration::from_millis(50)).await;
+        }
+
+        Self {
+            base_url,
+            client,
+            db_name,
+            mongodb_uri,
+        }
+    }
+
+    /// GET helper
+    pub async fn get(&self, path: &str) -> reqwest::Response {
+        self.client
+            .get(format!("{}{path}", self.base_url))
+            .send()
+            .await
+            .expect("GET request failed")
+    }
+
+    /// POST helper with JSON body
+    pub async fn post(&self, path: &str, body: &serde_json::Value) -> reqwest::Response {
+        self.client
+            .post(format!("{}{path}", self.base_url))
+            .json(body)
+            .send()
+            .await
+            .expect("POST request failed")
+    }
+
+    /// PATCH helper with JSON body
+    pub async fn patch(&self, path: &str, body: &serde_json::Value) -> reqwest::Response {
+        self.client
+            .patch(format!("{}{path}", self.base_url))
+            .json(body)
+            .send()
+            .await
+            .expect("PATCH request failed")
+    }
+
+    /// DELETE helper
+    pub async fn delete(&self, path: &str) -> reqwest::Response {
+        self.client
+            .delete(format!("{}{path}", self.base_url))
+            .send()
+            .await
+            .expect("DELETE request failed")
+    }
+
+    /// Get the unique database name for direct MongoDB access in tests.
+    pub fn db_name(&self) -> &str {
+        &self.db_name
+    }
+
+    /// Drop every per-tenant database belonging to this test run.
+    /// Post-M7.2-D the agent never opens a `db_name` directly —
+    /// data lives only in `<db_name>_<tenant>` per-tenant databases.
+    pub async fn cleanup(&self) {
+        if let Ok(client) = mongodb::Client::with_uri_str(&self.mongodb_uri).await {
+            if let Ok(names) = client.list_database_names().await {
+                let prefix = format!("{}_", self.db_name);
+                for name in names {
+                    if name.starts_with(&prefix) {
+                        client.database(&name).drop().await.ok();
+                    }
+                }
+            }
+        }
+    }
+}

compliance-agent/tests/e2e.rs

@@ -0,0 +1,7 @@
+// E2E test entry point.
+//
+// Run with:   cargo test -p compliance-agent --test e2e
+// Requires:   MongoDB running (set TEST_MONGODB_URI if not default)
+
+mod common;
+mod integration;

compliance-agent/tests/integration/api/cascade_delete.rs

@@ -0,0 +1,221 @@
+use crate::common::TestServer;
+use serde_json::json;
+
+/// Insert a DAST target directly into MongoDB linked to a repo.
+async fn insert_dast_target(server: &TestServer, repo_id: &str, name: &str) -> String {
+    let mongodb_uri = std::env::var("TEST_MONGODB_URI")
+        .unwrap_or_else(|_| "mongodb://root:example@localhost:27017/?authSource=admin".into());
+    let client = mongodb::Client::with_uri_str(&mongodb_uri).await.unwrap();
+    let db = client.database(&server.db_name());
+
+    let result = db
+        .collection::<mongodb::bson::Document>("dast_targets")
+        .insert_one(mongodb::bson::doc! {
+            "name": name,
+            "base_url": format!("https://{name}.example.com"),
+            "target_type": "webapp",
+            "repo_id": repo_id,
+            "rate_limit": 10,
+            "allow_destructive": false,
+            "created_at": mongodb::bson::DateTime::now(),
+        })
+        .await
+        .unwrap();
+
+    result.inserted_id.as_object_id().unwrap().to_hex()
+}
+
+/// Insert a pentest session linked to a target.
+async fn insert_pentest_session(server: &TestServer, target_id: &str, repo_id: &str) -> String {
+    let mongodb_uri = std::env::var("TEST_MONGODB_URI")
+        .unwrap_or_else(|_| "mongodb://root:example@localhost:27017/?authSource=admin".into());
+    let client = mongodb::Client::with_uri_str(&mongodb_uri).await.unwrap();
+    let db = client.database(&server.db_name());
+
+    let result = db
+        .collection::<mongodb::bson::Document>("pentest_sessions")
+        .insert_one(mongodb::bson::doc! {
+            "target_id": target_id,
+            "repo_id": repo_id,
+            "strategy": "comprehensive",
+            "status": "completed",
+            "findings_count": 1_i32,
+            "exploitable_count": 0_i32,
+            "created_at": mongodb::bson::DateTime::now(),
+        })
+        .await
+        .unwrap();
+
+    result.inserted_id.as_object_id().unwrap().to_hex()
+}
+
+/// Insert an attack chain node linked to a session.
+async fn insert_attack_node(server: &TestServer, session_id: &str) {
+    let mongodb_uri = std::env::var("TEST_MONGODB_URI")
+        .unwrap_or_else(|_| "mongodb://root:example@localhost:27017/?authSource=admin".into());
+    let client = mongodb::Client::with_uri_str(&mongodb_uri).await.unwrap();
+    let db = client.database(&server.db_name());
+
+    db.collection::<mongodb::bson::Document>("attack_chain_nodes")
+        .insert_one(mongodb::bson::doc! {
+            "session_id": session_id,
+            "node_id": "node-1",
+            "tool_name": "recon",
+            "status": "completed",
+            "created_at": mongodb::bson::DateTime::now(),
+        })
+        .await
+        .unwrap();
+}
+
+/// Insert a DAST finding linked to a target.
+async fn insert_dast_finding(server: &TestServer, target_id: &str, session_id: &str) {
+    let mongodb_uri = std::env::var("TEST_MONGODB_URI")
+        .unwrap_or_else(|_| "mongodb://root:example@localhost:27017/?authSource=admin".into());
+    let client = mongodb::Client::with_uri_str(&mongodb_uri).await.unwrap();
+    let db = client.database(&server.db_name());
+
+    db.collection::<mongodb::bson::Document>("dast_findings")
+        .insert_one(mongodb::bson::doc! {
+            "scan_run_id": "run-1",
+            "target_id": target_id,
+            "vuln_type": "xss",
+            "title": "Reflected XSS",
+            "description": "XSS in search param",
+            "severity": "high",
+            "endpoint": "https://example.com/search",
+            "method": "GET",
+            "exploitable": true,
+            "evidence": [],
+            "session_id": session_id,
+            "created_at": mongodb::bson::DateTime::now(),
+        })
+        .await
+        .unwrap();
+}
+
+/// Helper to count documents in a collection
+async fn count_docs(server: &TestServer, collection: &str) -> u64 {
+    let mongodb_uri = std::env::var("TEST_MONGODB_URI")
+        .unwrap_or_else(|_| "mongodb://root:example@localhost:27017/?authSource=admin".into());
+    let client = mongodb::Client::with_uri_str(&mongodb_uri).await.unwrap();
+    let db = client.database(&server.db_name());
+    db.collection::<mongodb::bson::Document>(collection)
+        .count_documents(mongodb::bson::doc! {})
+        .await
+        .unwrap()
+}
+
+#[tokio::test]
+async fn delete_repo_cascades_to_dast_and_pentest_data() {
+    let server = TestServer::start().await;
+
+    // Create a repo
+    let resp = server
+        .post(
+            "/api/v1/repositories",
+            &json!({
+                "name": "cascade-test",
+                "git_url": "https://github.com/example/cascade-test.git",
+            }),
+        )
+        .await;
+    let body: serde_json::Value = resp.json().await.unwrap();
+    let repo_id = body["data"]["id"].as_str().unwrap().to_string();
+
+    // Insert DAST target linked to repo
+    let target_id = insert_dast_target(&server, &repo_id, "cascade-target").await;
+
+    // Insert pentest session linked to target
+    let session_id = insert_pentest_session(&server, &target_id, &repo_id).await;
+
+    // Insert downstream data
+    insert_attack_node(&server, &session_id).await;
+    insert_dast_finding(&server, &target_id, &session_id).await;
+
+    // Verify data exists
+    assert_eq!(count_docs(&server, "dast_targets").await, 1);
+    assert_eq!(count_docs(&server, "pentest_sessions").await, 1);
+    assert_eq!(count_docs(&server, "attack_chain_nodes").await, 1);
+    assert_eq!(count_docs(&server, "dast_findings").await, 1);
+
+    // Delete the repo
+    let resp = server
+        .delete(&format!("/api/v1/repositories/{repo_id}"))
+        .await;
+    assert_eq!(resp.status(), 200);
+
+    // All downstream data should be gone
+    assert_eq!(count_docs(&server, "dast_targets").await, 0);
+    assert_eq!(count_docs(&server, "pentest_sessions").await, 0);
+    assert_eq!(count_docs(&server, "attack_chain_nodes").await, 0);
+    assert_eq!(count_docs(&server, "dast_findings").await, 0);
+
+    server.cleanup().await;
+}
+
+#[tokio::test]
+async fn delete_repo_cascades_sast_findings_and_sbom() {
+    let server = TestServer::start().await;
+
+    // Create a repo
+    let resp = server
+        .post(
+            "/api/v1/repositories",
+            &json!({
+                "name": "sast-cascade",
+                "git_url": "https://github.com/example/sast-cascade.git",
+            }),
+        )
+        .await;
+    let body: serde_json::Value = resp.json().await.unwrap();
+    let repo_id = body["data"]["id"].as_str().unwrap().to_string();
+
+    // Insert SAST finding and SBOM entry
+    let mongodb_uri = std::env::var("TEST_MONGODB_URI")
+        .unwrap_or_else(|_| "mongodb://root:example@localhost:27017/?authSource=admin".into());
+    let client = mongodb::Client::with_uri_str(&mongodb_uri).await.unwrap();
+    let db = client.database(&server.db_name());
+    let now = mongodb::bson::DateTime::now();
+
+    db.collection::<mongodb::bson::Document>("findings")
+        .insert_one(mongodb::bson::doc! {
+            "repo_id": &repo_id,
+            "fingerprint": "fp-test-1",
+            "scanner": "semgrep",
+            "scan_type": "sast",
+            "title": "SQL Injection",
+            "description": "desc",
+            "severity": "critical",
+            "status": "open",
+            "created_at": now,
+            "updated_at": now,
+        })
+        .await
+        .unwrap();
+
+    db.collection::<mongodb::bson::Document>("sbom_entries")
+        .insert_one(mongodb::bson::doc! {
+            "repo_id": &repo_id,
+            "name": "lodash",
+            "version": "4.17.20",
+            "package_manager": "npm",
+            "known_vulnerabilities": [],
+        })
+        .await
+        .unwrap();
+
+    assert_eq!(count_docs(&server, "findings").await, 1);
+    assert_eq!(count_docs(&server, "sbom_entries").await, 1);
+
+    // Delete repo
+    server
+        .delete(&format!("/api/v1/repositories/{repo_id}"))
+        .await;
+
+    // Both should be gone
+    assert_eq!(count_docs(&server, "findings").await, 0);
+    assert_eq!(count_docs(&server, "sbom_entries").await, 0);
+
+    server.cleanup().await;
+}

compliance-agent/tests/integration/api/dast.rs

@@ -0,0 +1,48 @@
+use crate::common::TestServer;
+use serde_json::json;
+
+#[tokio::test]
+async fn add_and_list_dast_targets() {
+    let server = TestServer::start().await;
+
+    // Initially empty
+    let resp = server.get("/api/v1/dast/targets").await;
+    assert_eq!(resp.status(), 200);
+    let body: serde_json::Value = resp.json().await.unwrap();
+    assert_eq!(body["data"].as_array().unwrap().len(), 0);
+
+    // Add a target
+    let resp = server
+        .post(
+            "/api/v1/dast/targets",
+            &json!({
+                "name": "test-app",
+                "base_url": "https://test-app.example.com",
+                "target_type": "webapp",
+            }),
+        )
+        .await;
+    assert_eq!(resp.status(), 200);
+
+    // List should return 1
+    let resp = server.get("/api/v1/dast/targets").await;
+    let body: serde_json::Value = resp.json().await.unwrap();
+    let targets = body["data"].as_array().unwrap();
+    assert_eq!(targets.len(), 1);
+    assert_eq!(targets[0]["name"], "test-app");
+    assert_eq!(targets[0]["base_url"], "https://test-app.example.com");
+
+    server.cleanup().await;
+}
+
+#[tokio::test]
+async fn list_dast_findings_empty() {
+    let server = TestServer::start().await;
+
+    let resp = server.get("/api/v1/dast/findings").await;
+    assert_eq!(resp.status(), 200);
+    let body: serde_json::Value = resp.json().await.unwrap();
+    assert_eq!(body["data"].as_array().unwrap().len(), 0);
+
+    server.cleanup().await;
+}

compliance-agent/tests/integration/api/findings.rs

@@ -0,0 +1,144 @@
+use crate::common::TestServer;
+use serde_json::json;
+
+/// Helper: insert a finding directly via MongoDB for testing query endpoints.
+async fn insert_finding(server: &TestServer, repo_id: &str, title: &str, severity: &str) {
+    // We insert via the agent's DB by posting to the internal test path.
+    // Since there's no direct "create finding" API, we use MongoDB directly.
+    let mongodb_uri = std::env::var("TEST_MONGODB_URI")
+        .unwrap_or_else(|_| "mongodb://root:example@localhost:27017/?authSource=admin".into());
+
+    // Extract the database name from the server's unique DB
+    // We'll use the agent's internal DB through the stats endpoint to verify
+    let client = mongodb::Client::with_uri_str(&mongodb_uri).await.unwrap();
+
+    // Get the DB name from the test server by parsing the health response
+    // For now, we use a direct insert approach
+    let db = client.database(&server.db_name());
+
+    let now = mongodb::bson::DateTime::now();
+    db.collection::<mongodb::bson::Document>("findings")
+        .insert_one(mongodb::bson::doc! {
+            "repo_id": repo_id,
+            "fingerprint": format!("fp-{title}-{severity}"),
+            "scanner": "test-scanner",
+            "scan_type": "sast",
+            "title": title,
+            "description": format!("Test finding: {title}"),
+            "severity": severity,
+            "status": "open",
+            "created_at": now,
+            "updated_at": now,
+        })
+        .await
+        .unwrap();
+}
+
+#[tokio::test]
+async fn list_findings_empty() {
+    let server = TestServer::start().await;
+
+    let resp = server.get("/api/v1/findings").await;
+    assert_eq!(resp.status(), 200);
+
+    let body: serde_json::Value = resp.json().await.unwrap();
+    assert_eq!(body["data"].as_array().unwrap().len(), 0);
+    assert_eq!(body["total"], 0);
+
+    server.cleanup().await;
+}
+
+#[tokio::test]
+async fn list_findings_with_data() {
+    let server = TestServer::start().await;
+
+    insert_finding(&server, "repo1", "SQL Injection", "critical").await;
+    insert_finding(&server, "repo1", "XSS", "high").await;
+    insert_finding(&server, "repo2", "Info Leak", "low").await;
+
+    let resp = server.get("/api/v1/findings").await;
+    assert_eq!(resp.status(), 200);
+    let body: serde_json::Value = resp.json().await.unwrap();
+    assert_eq!(body["total"], 3);
+
+    // Filter by severity
+    let resp = server.get("/api/v1/findings?severity=critical").await;
+    let body: serde_json::Value = resp.json().await.unwrap();
+    assert_eq!(body["total"], 1);
+    assert_eq!(body["data"][0]["title"], "SQL Injection");
+
+    // Filter by repo
+    let resp = server.get("/api/v1/findings?repo_id=repo1").await;
+    let body: serde_json::Value = resp.json().await.unwrap();
+    assert_eq!(body["total"], 2);
+
+    server.cleanup().await;
+}
+
+#[tokio::test]
+async fn update_finding_status() {
+    let server = TestServer::start().await;
+
+    insert_finding(&server, "repo1", "Test Bug", "medium").await;
+
+    // Get the finding ID
+    let resp = server.get("/api/v1/findings").await;
+    let body: serde_json::Value = resp.json().await.unwrap();
+    let finding_id = body["data"][0]["_id"]["$oid"].as_str().unwrap();
+
+    // Update status to resolved
+    let resp = server
+        .patch(
+            &format!("/api/v1/findings/{finding_id}/status"),
+            &json!({ "status": "resolved" }),
+        )
+        .await;
+    assert_eq!(resp.status(), 200);
+
+    // Verify it's updated
+    let resp = server.get(&format!("/api/v1/findings/{finding_id}")).await;
+    assert_eq!(resp.status(), 200);
+    let body: serde_json::Value = resp.json().await.unwrap();
+    assert_eq!(body["data"]["status"], "resolved");
+
+    server.cleanup().await;
+}
+
+#[tokio::test]
+async fn bulk_update_finding_status() {
+    let server = TestServer::start().await;
+
+    insert_finding(&server, "repo1", "Bug A", "high").await;
+    insert_finding(&server, "repo1", "Bug B", "high").await;
+
+    // Get both finding IDs
+    let resp = server.get("/api/v1/findings").await;
+    let body: serde_json::Value = resp.json().await.unwrap();
+    let ids: Vec<String> = body["data"]
+        .as_array()
+        .unwrap()
+        .iter()
+        .map(|f| f["_id"]["$oid"].as_str().unwrap().to_string())
+        .collect();
+
+    // Bulk update
+    let resp = server
+        .patch(
+            "/api/v1/findings/bulk-status",
+            &json!({
+                "ids": ids,
+                "status": "false_positive"
+            }),
+        )
+        .await;
+    assert_eq!(resp.status(), 200);
+
+    // Verify both are updated
+    for id in &ids {
+        let resp = server.get(&format!("/api/v1/findings/{id}")).await;
+        let body: serde_json::Value = resp.json().await.unwrap();
+        assert_eq!(body["data"]["status"], "false_positive");
+    }
+
+    server.cleanup().await;
+}

compliance-agent/tests/integration/api/health.rs

@@ -0,0 +1,29 @@
+use crate::common::TestServer;
+
+#[tokio::test]
+async fn health_endpoint_returns_ok() {
+    let server = TestServer::start().await;
+
+    let resp = server.get("/api/v1/health").await;
+    assert_eq!(resp.status(), 200);
+
+    let body: serde_json::Value = resp.json().await.unwrap();
+    assert_eq!(body["status"], "ok");
+
+    server.cleanup().await;
+}
+
+#[tokio::test]
+async fn stats_overview_returns_zeroes_on_empty_db() {
+    let server = TestServer::start().await;
+
+    let resp = server.get("/api/v1/stats/overview").await;
+    assert_eq!(resp.status(), 200);
+
+    let body: serde_json::Value = resp.json().await.unwrap();
+    let data = &body["data"];
+    assert_eq!(data["repositories"], 0);
+    assert_eq!(data["total_findings"], 0);
+
+    server.cleanup().await;
+}

compliance-agent/tests/integration/api/mod.rs

@@ -0,0 +1,6 @@
+mod cascade_delete;
+mod dast;
+mod findings;
+mod health;
+mod repositories;
+mod stats;

compliance-agent/tests/integration/api/repositories.rs

@@ -0,0 +1,110 @@
+use crate::common::TestServer;
+use serde_json::json;
+
+#[tokio::test]
+async fn add_and_list_repository() {
+    let server = TestServer::start().await;
+
+    // Initially empty
+    let resp = server.get("/api/v1/repositories").await;
+    assert_eq!(resp.status(), 200);
+    let body: serde_json::Value = resp.json().await.unwrap();
+    assert_eq!(body["data"].as_array().unwrap().len(), 0);
+
+    // Add a repository
+    let resp = server
+        .post(
+            "/api/v1/repositories",
+            &json!({
+                "name": "test-repo",
+                "git_url": "https://github.com/example/test-repo.git",
+            }),
+        )
+        .await;
+    assert_eq!(resp.status(), 200);
+    let body: serde_json::Value = resp.json().await.unwrap();
+    let repo_id = body["data"]["id"].as_str().unwrap().to_string();
+    assert!(!repo_id.is_empty());
+
+    // List should now return 1
+    let resp = server.get("/api/v1/repositories").await;
+    let body: serde_json::Value = resp.json().await.unwrap();
+    let repos = body["data"].as_array().unwrap();
+    assert_eq!(repos.len(), 1);
+    assert_eq!(repos[0]["name"], "test-repo");
+
+    server.cleanup().await;
+}
+
+#[tokio::test]
+async fn add_duplicate_repository_fails() {
+    let server = TestServer::start().await;
+
+    let payload = json!({
+        "name": "dup-repo",
+        "git_url": "https://github.com/example/dup-repo.git",
+    });
+
+    // First add succeeds
+    let resp = server.post("/api/v1/repositories", &payload).await;
+    assert_eq!(resp.status(), 200);
+
+    // Second add with same git_url should fail (unique index)
+    let resp = server.post("/api/v1/repositories", &payload).await;
+    assert_ne!(resp.status(), 200);
+
+    server.cleanup().await;
+}
+
+#[tokio::test]
+async fn delete_repository() {
+    let server = TestServer::start().await;
+
+    // Add a repo
+    let resp = server
+        .post(
+            "/api/v1/repositories",
+            &json!({
+                "name": "to-delete",
+                "git_url": "https://github.com/example/to-delete.git",
+            }),
+        )
+        .await;
+    let body: serde_json::Value = resp.json().await.unwrap();
+    let repo_id = body["data"]["id"].as_str().unwrap();
+
+    // Delete it
+    let resp = server
+        .delete(&format!("/api/v1/repositories/{repo_id}"))
+        .await;
+    assert_eq!(resp.status(), 200);
+
+    // List should be empty again
+    let resp = server.get("/api/v1/repositories").await;
+    let body: serde_json::Value = resp.json().await.unwrap();
+    assert_eq!(body["data"].as_array().unwrap().len(), 0);
+
+    server.cleanup().await;
+}
+
+#[tokio::test]
+async fn delete_nonexistent_repository_returns_404() {
+    let server = TestServer::start().await;
+
+    let resp = server
+        .delete("/api/v1/repositories/000000000000000000000000")
+        .await;
+    assert_eq!(resp.status(), 404);
+
+    server.cleanup().await;
+}
+
+#[tokio::test]
+async fn delete_invalid_id_returns_400() {
+    let server = TestServer::start().await;
+
+    let resp = server.delete("/api/v1/repositories/not-a-valid-id").await;
+    assert_eq!(resp.status(), 400);
+
+    server.cleanup().await;
+}

compliance-agent/tests/integration/api/stats.rs

@@ -0,0 +1,111 @@
+use crate::common::TestServer;
+use serde_json::json;
+
+#[tokio::test]
+async fn stats_overview_reflects_inserted_data() {
+    let server = TestServer::start().await;
+
+    // Add a repo
+    server
+        .post(
+            "/api/v1/repositories",
+            &json!({
+                "name": "stats-repo",
+                "git_url": "https://github.com/example/stats-repo.git",
+            }),
+        )
+        .await;
+
+    // Insert findings directly
+    let mongodb_uri = std::env::var("TEST_MONGODB_URI")
+        .unwrap_or_else(|_| "mongodb://root:example@localhost:27017/?authSource=admin".into());
+    let client = mongodb::Client::with_uri_str(&mongodb_uri).await.unwrap();
+    let db = client.database(&server.db_name());
+    let now = mongodb::bson::DateTime::now();
+
+    for (title, severity) in [
+        ("Critical Bug", "critical"),
+        ("High Bug", "high"),
+        ("Medium Bug", "medium"),
+        ("Low Bug", "low"),
+    ] {
+        db.collection::<mongodb::bson::Document>("findings")
+            .insert_one(mongodb::bson::doc! {
+                "repo_id": "test-repo-id",
+                "fingerprint": format!("fp-{title}"),
+                "scanner": "test",
+                "scan_type": "sast",
+                "title": title,
+                "description": "desc",
+                "severity": severity,
+                "status": "open",
+                "created_at": now,
+                "updated_at": now,
+            })
+            .await
+            .unwrap();
+    }
+
+    let resp = server.get("/api/v1/stats/overview").await;
+    assert_eq!(resp.status(), 200);
+
+    let body: serde_json::Value = resp.json().await.unwrap();
+    let data = &body["data"];
+    assert_eq!(data["repositories"], 1);
+    assert_eq!(data["total_findings"], 4);
+    assert_eq!(data["critical"], 1);
+    assert_eq!(data["high"], 1);
+
+    server.cleanup().await;
+}
+
+#[tokio::test]
+async fn stats_update_after_finding_status_change() {
+    let server = TestServer::start().await;
+
+    // Insert a finding
+    let mongodb_uri = std::env::var("TEST_MONGODB_URI")
+        .unwrap_or_else(|_| "mongodb://root:example@localhost:27017/?authSource=admin".into());
+    let client = mongodb::Client::with_uri_str(&mongodb_uri).await.unwrap();
+    let db = client.database(&server.db_name());
+    let now = mongodb::bson::DateTime::now();
+
+    let result = db
+        .collection::<mongodb::bson::Document>("findings")
+        .insert_one(mongodb::bson::doc! {
+            "repo_id": "repo-1",
+            "fingerprint": "fp-stats-test",
+            "scanner": "test",
+            "scan_type": "sast",
+            "title": "Stats Test Finding",
+            "description": "desc",
+            "severity": "high",
+            "status": "open",
+            "created_at": now,
+            "updated_at": now,
+        })
+        .await
+        .unwrap();
+    let finding_id = result.inserted_id.as_object_id().unwrap().to_hex();
+
+    // Stats should show 1 finding
+    let resp = server.get("/api/v1/stats/overview").await;
+    let body: serde_json::Value = resp.json().await.unwrap();
+    assert_eq!(body["data"]["total_findings"], 1);
+
+    // Mark it as resolved
+    server
+        .patch(
+            &format!("/api/v1/findings/{finding_id}/status"),
+            &json!({ "status": "resolved" }),
+        )
+        .await;
+
+    // The finding still exists (status changed, not deleted)
+    let resp = server.get("/api/v1/stats/overview").await;
+    let body: serde_json::Value = resp.json().await.unwrap();
+    // total_findings counts all findings regardless of status
+    assert_eq!(body["data"]["total_findings"], 1);
+
+    server.cleanup().await;
+}

compliance-agent/tests/integration/mod.rs

@@ -1,4 +1,9 @@
-// Integration tests for the compliance-agent crate.
+// E2E / Integration tests for the compliance-agent API.
 //
-// Add tests that exercise the full pipeline, API handlers,
-// and cross-module interactions here.
+// These tests require a running MongoDB instance. Set TEST_MONGODB_URI
+// if it's not at the default `mongodb://root:example@localhost:27017`.
+//
+// Run with:   cargo test -p compliance-agent --test e2e
+// Or nightly: (via CI with MongoDB service container)
+
+mod api;

compliance-agent/tests/tenant_isolation.rs

@@ -0,0 +1,298 @@
+//! M7.2-A — `DatabasePool` isolation proof.
+//!
+//! Two `TenantContext`s, two databases, one client. Insert on A, query
+//! on B → empty. Insert on B, query on A → only A's docs. Proves that
+//! the per-tenant database split actually isolates at the driver level
+//! and not at "we hope we filter."
+//!
+//! Requires MongoDB. Set `TEST_MONGODB_URI` to override the default
+//! `mongodb://root:example@localhost:27017/?authSource=admin`.
+
+#![allow(clippy::expect_used, clippy::unwrap_used)]
+
+use compliance_agent::database::DatabasePool;
+use compliance_core::models::TrackedRepository;
+use compliance_core::{OrgRole, TenantContext, TenantStatus};
+use mongodb::bson::doc;
+
+fn ctx(tenant_id: &str, slug: &str) -> TenantContext {
+    TenantContext {
+        tenant_id: tenant_id.to_string(),
+        tenant_slug: slug.to_string(),
+        org_roles: vec![OrgRole::ItAdmin],
+        products: vec!["compliance-scanner".to_string()],
+        plan: "starter".to_string(),
+        status: TenantStatus::Active,
+        user_id: "u-1".to_string(),
+        user_name: None,
+    }
+}
+
+fn fixture_repo(name: &str, git_url: &str) -> TrackedRepository {
+    TrackedRepository {
+        id: None,
+        name: name.to_string(),
+        git_url: git_url.to_string(),
+        default_branch: "main".to_string(),
+        local_path: None,
+        scan_schedule: None,
+        webhook_enabled: false,
+        webhook_secret: None,
+        tracker_type: None,
+        tracker_owner: None,
+        tracker_repo: None,
+        tracker_token: None,
+        auth_token: None,
+        auth_username: None,
+        last_scanned_commit: None,
+        findings_count: 0,
+        created_at: chrono::Utc::now(),
+        updated_at: chrono::Utc::now(),
+    }
+}
+
+#[tokio::test]
+async fn pool_isolates_tenants_at_driver_level() {
+    let uri = std::env::var("TEST_MONGODB_URI")
+        .unwrap_or_else(|_| "mongodb://root:example@localhost:27017/?authSource=admin".into());
+    // Unique per run so parallel test invocations don't collide. Kept
+    // short because Mongo caps db names at 63 bytes (prefix + tenant_id).
+    let prefix = format!("m72a_{}", short_id());
+
+    let pool = DatabasePool::connect(&uri, &prefix)
+        .await
+        .expect("Failed to connect to MongoDB — is it running?");
+
+    let acme = ctx("00000000-0000-0000-0000-00000000acme", "acme");
+    let globex = ctx("00000000-0000-0000-0000-0000globex000", "globex");
+
+    let acme_db = pool.for_tenant(&acme).await.expect("acme db");
+    let globex_db = pool.for_tenant(&globex).await.expect("globex db");
+
+    // Write distinct repos into each tenant's database.
+    acme_db
+        .repositories()
+        .insert_one(fixture_repo("acme-app", "git@example.com:acme/app.git"))
+        .await
+        .expect("insert acme");
+    globex_db
+        .repositories()
+        .insert_one(fixture_repo(
+            "globex-platform",
+            "git@example.com:globex/platform.git",
+        ))
+        .await
+        .expect("insert globex");
+
+    // The point of the whole exercise: acme can ONLY see acme's repo
+    // and globex can ONLY see globex's, with no filter doc anywhere
+    // because the isolation is at the database handle, not in the query.
+    let acme_seen = collect(&acme_db).await;
+    let globex_seen = collect(&globex_db).await;
+
+    assert_eq!(acme_seen.len(), 1, "acme should see exactly its own repo");
+    assert_eq!(acme_seen[0].name, "acme-app");
+    assert_eq!(
+        globex_seen.len(),
+        1,
+        "globex should see exactly its own repo"
+    );
+    assert_eq!(globex_seen[0].name, "globex-platform");
+
+    // Sanity: the two databases really are different by name.
+    let acme_db_name = pool.tenant_db_name(&acme.tenant_id);
+    let globex_db_name = pool.tenant_db_name(&globex.tenant_id);
+    assert_ne!(acme_db_name, globex_db_name);
+    assert!(acme_db_name.starts_with(&prefix));
+
+    // Cleanup — drop both per-tenant databases.
+    pool.client()
+        .database(&acme_db_name)
+        .drop()
+        .await
+        .expect("drop acme");
+    pool.client()
+        .database(&globex_db_name)
+        .drop()
+        .await
+        .expect("drop globex");
+}
+
+#[tokio::test]
+async fn for_tenant_is_idempotent_index_creation() {
+    let uri = std::env::var("TEST_MONGODB_URI")
+        .unwrap_or_else(|_| "mongodb://root:example@localhost:27017/?authSource=admin".into());
+    let prefix = format!("m72a_{}", short_id());
+    let pool = DatabasePool::connect(&uri, &prefix).await.expect("connect");
+
+    let acme = ctx("00000000-0000-0000-0000-00000000acme", "acme");
+
+    // Second call must not fail (ensure_indexes already ran, in-memory
+    // marker is set, Mongo's createIndex is idempotent by name anyway).
+    let _ = pool.for_tenant(&acme).await.expect("first call");
+    let _ = pool.for_tenant(&acme).await.expect("second call");
+    let _ = pool.for_tenant(&acme).await.expect("third call");
+
+    // Cleanup
+    let db_name = pool.tenant_db_name(&acme.tenant_id);
+    pool.client().database(&db_name).drop().await.expect("drop");
+}
+
+#[tokio::test]
+async fn tenant_db_name_sanitizes_unsafe_characters() {
+    let uri = std::env::var("TEST_MONGODB_URI")
+        .unwrap_or_else(|_| "mongodb://root:example@localhost:27017/?authSource=admin".into());
+    let pool = DatabasePool::connect(&uri, "m72a_sanitize")
+        .await
+        .expect("connect");
+
+    // Mongo db names cannot contain `/ \ . " $ <space> NUL`. The pool
+    // must rewrite these without exploding on connect.
+    let funky = "te/n.a\\nt$id\" with spaces";
+    let name = pool.tenant_db_name(funky);
+    for c in ['/', '\\', '.', '"', '$', ' '] {
+        assert!(
+            !name.contains(c),
+            "sanitized db name still contains {c:?}: {name}"
+        );
+    }
+}
+
+#[tokio::test]
+async fn admin_helpers_list_and_drop_tenant_dbs() {
+    let uri = std::env::var("TEST_MONGODB_URI")
+        .unwrap_or_else(|_| "mongodb://root:example@localhost:27017/?authSource=admin".into());
+    let prefix = format!("m72d_{}", short_id());
+    let pool = DatabasePool::connect(&uri, &prefix).await.expect("connect");
+
+    let acme = ctx("00000000-0000-0000-0000-00000000acme", "acme");
+    let globex = ctx("00000000-0000-0000-0000-0000globex000", "globex");
+
+    // Provision two tenants and write a doc into each so the databases
+    // actually materialize on the cluster (Mongo lazily creates DBs).
+    let acme_db = pool.for_tenant(&acme).await.expect("acme db");
+    let globex_db = pool.for_tenant(&globex).await.expect("globex db");
+    acme_db
+        .repositories()
+        .insert_one(fixture_repo("acme-app", "git@example.com:acme/app.git"))
+        .await
+        .expect("insert acme");
+    globex_db
+        .repositories()
+        .insert_one(fixture_repo("globex-app", "git@example.com:globex/app.git"))
+        .await
+        .expect("insert globex");
+
+    // list_tenant_db_names sees both, filtered by prefix
+    let names = pool.list_tenant_db_names().await.expect("list tenants");
+    let acme_name = pool.tenant_db_name(&acme.tenant_id);
+    let globex_name = pool.tenant_db_name(&globex.tenant_id);
+    assert!(
+        names.contains(&acme_name),
+        "expected {acme_name} in {names:?}"
+    );
+    assert!(
+        names.contains(&globex_name),
+        "expected {globex_name} in {names:?}"
+    );
+    for name in &names {
+        assert!(name.starts_with(&format!("{prefix}_")));
+    }
+
+    // drop_tenant removes acme's DB
+    pool.drop_tenant(&acme.tenant_id)
+        .await
+        .expect("drop acme tenant");
+    let after = pool
+        .list_tenant_db_names()
+        .await
+        .expect("list tenants after drop");
+    assert!(
+        !after.contains(&acme_name),
+        "acme should be gone after drop, got {after:?}"
+    );
+    assert!(
+        after.contains(&globex_name),
+        "globex should still be present, got {after:?}"
+    );
+
+    // Cleanup remaining
+    pool.drop_tenant(&globex.tenant_id)
+        .await
+        .expect("drop globex tenant");
+}
+
+#[tokio::test]
+async fn tenant_db_name_falls_back_to_hash_when_too_long() {
+    let uri = std::env::var("TEST_MONGODB_URI")
+        .unwrap_or_else(|_| "mongodb://root:example@localhost:27017/?authSource=admin".into());
+    let pool = DatabasePool::connect(&uri, "m72a_long")
+        .await
+        .expect("connect");
+
+    // 100-byte tenant_id would overflow the 63-byte db-name cap with
+    // any reasonable prefix. The pool must hash it down.
+    let huge = "x".repeat(100);
+    let name = pool.tenant_db_name(&huge);
+    assert!(name.len() <= 63, "hashed name should fit: {name}");
+    assert!(name.starts_with("m72a_long_"));
+    // The hash suffix is 32 hex chars (16-byte SHA-256 truncation).
+    let suffix = name.trim_start_matches("m72a_long_");
+    assert_eq!(
+        suffix.len(),
+        32,
+        "expected 32-hex suffix (16-byte hash), got {suffix:?}"
+    );
+    assert!(suffix.chars().all(|c| c.is_ascii_hexdigit()));
+
+    // Stable: same input → same output.
+    assert_eq!(name, pool.tenant_db_name(&huge));
+
+    // Different inputs → different outputs (collision check on a tiny
+    // sample — full birthday-resistance is a proof not a test).
+    let huge2 = "y".repeat(100);
+    assert_ne!(pool.tenant_db_name(&huge), pool.tenant_db_name(&huge2));
+}
+
+#[tokio::test]
+async fn connect_rejects_overlong_db_prefix() {
+    let uri = std::env::var("TEST_MONGODB_URI")
+        .unwrap_or_else(|_| "mongodb://root:example@localhost:27017/?authSource=admin".into());
+
+    // MAX_PREFIX_LEN is 30 (= 63 - 1 - 32). A 31-char prefix MUST be
+    // rejected at construction so the hash-fallback path can never
+    // produce an over-long db name at runtime.
+    let too_long = "a".repeat(31);
+    let err = DatabasePool::connect(&uri, &too_long).await.unwrap_err();
+    let msg = format!("{err}");
+    assert!(
+        msg.contains("max is 30") || msg.contains(&too_long),
+        "error should explain the cap: {msg}"
+    );
+
+    // Exactly 30 chars is the inclusive bound — must succeed.
+    let just_right = "a".repeat(30);
+    let _ = DatabasePool::connect(&uri, &just_right)
+        .await
+        .expect("30-char prefix should be accepted");
+}
+
+/// Short UUID slug for keeping test prefixes well under Mongo's 63-byte
+/// db-name cap.
+fn short_id() -> String {
+    uuid::Uuid::new_v4().simple().to_string()[..8].to_string()
+}
+
+/// Drain a `repositories` find cursor on the given tenant database.
+async fn collect(db: &compliance_agent::database::Database) -> Vec<TrackedRepository> {
+    let mut cursor = db
+        .repositories()
+        .find(doc! {})
+        .await
+        .expect("find repositories");
+    let mut out = Vec::new();
+    while cursor.advance().await.expect("advance") {
+        out.push(cursor.deserialize_current().expect("deserialize"));
+    }
+    out
+}

compliance-agent/tests/tenant_status_middleware.rs

@@ -0,0 +1,122 @@
+//! M7.1 — integration tests for `compliance_core::auth::require_tenant_status`.
+//!
+//! Exercises the middleware end-to-end through an Axum router so we
+//! catch wiring bugs (extension propagation, method matching) that pure
+//! unit tests would miss.
+
+#![allow(clippy::expect_used, clippy::unwrap_used)]
+
+use axum::{
+    body::Body,
+    extract::Request,
+    http::{Method, StatusCode},
+    middleware::{from_fn, Next},
+    response::Response,
+    routing::{get, post},
+    Router,
+};
+use compliance_core::{auth::require_tenant_status, TenantContext, TenantStatus};
+use tower::ServiceExt;
+
+fn ctx_with(status: TenantStatus) -> TenantContext {
+    TenantContext {
+        tenant_id: "t-1".to_string(),
+        tenant_slug: "acme".to_string(),
+        org_roles: vec![],
+        products: vec![],
+        plan: "starter".to_string(),
+        status,
+        user_id: "u-1".to_string(),
+        user_name: None,
+    }
+}
+
+fn router_with_ctx(ctx: Option<TenantContext>) -> Router {
+    let injector = move |mut req: Request, next: Next| {
+        let ctx = ctx.clone();
+        async move {
+            if let Some(c) = ctx {
+                req.extensions_mut().insert(c);
+            }
+            next.run(req).await
+        }
+    };
+
+    Router::new()
+        .route("/r", get(|| async { "read" }))
+        .route("/w", post(|| async { "write" }))
+        .layer(from_fn(require_tenant_status))
+        .layer(from_fn(injector))
+}
+
+async fn call(router: Router, method: Method, path: &str) -> Response {
+    let req = Request::builder()
+        .method(method)
+        .uri(path)
+        .body(Body::empty())
+        .expect("request build");
+    router.oneshot(req).await.expect("oneshot")
+}
+
+#[tokio::test]
+async fn active_tenant_can_read_and_write() {
+    let r = router_with_ctx(Some(ctx_with(TenantStatus::Active)));
+    assert_eq!(
+        call(r.clone(), Method::GET, "/r").await.status(),
+        StatusCode::OK
+    );
+    assert_eq!(call(r, Method::POST, "/w").await.status(), StatusCode::OK);
+}
+
+#[tokio::test]
+async fn trial_tenant_can_read_and_write() {
+    let r = router_with_ctx(Some(ctx_with(TenantStatus::Trial)));
+    assert_eq!(
+        call(r.clone(), Method::GET, "/r").await.status(),
+        StatusCode::OK
+    );
+    assert_eq!(call(r, Method::POST, "/w").await.status(), StatusCode::OK);
+}
+
+#[tokio::test]
+async fn demo_tenant_can_read_and_write() {
+    let r = router_with_ctx(Some(ctx_with(TenantStatus::Demo)));
+    assert_eq!(
+        call(r.clone(), Method::GET, "/r").await.status(),
+        StatusCode::OK
+    );
+    assert_eq!(call(r, Method::POST, "/w").await.status(), StatusCode::OK);
+}
+
+#[tokio::test]
+async fn frozen_tenant_can_read_but_not_write() {
+    let r = router_with_ctx(Some(ctx_with(TenantStatus::Frozen)));
+    assert_eq!(
+        call(r.clone(), Method::GET, "/r").await.status(),
+        StatusCode::OK
+    );
+    assert_eq!(
+        call(r, Method::POST, "/w").await.status(),
+        StatusCode::PAYMENT_REQUIRED
+    );
+}
+
+#[tokio::test]
+async fn archived_tenant_is_gone_on_every_method() {
+    let r = router_with_ctx(Some(ctx_with(TenantStatus::Archived)));
+    assert_eq!(
+        call(r.clone(), Method::GET, "/r").await.status(),
+        StatusCode::GONE
+    );
+    assert_eq!(call(r, Method::POST, "/w").await.status(), StatusCode::GONE);
+}
+
+#[tokio::test]
+async fn no_context_passes_through() {
+    let r = router_with_ctx(None);
+    assert_eq!(
+        call(r.clone(), Method::GET, "/r").await.status(),
+        StatusCode::OK
+    );
+    assert_eq!(call(r, Method::POST, "/w").await.status(), StatusCode::OK);
+}

compliance-core/Cargo.toml

@@ -18,6 +18,15 @@ telemetry = [
     "dep:tracing-subscriber",
     "dep:tracing",
 ]
+# Pulls in the M7.1 Axum middleware + extractor. Consumers that don't
+# embed an HTTP server (e.g. the wasm dashboard frontend) leave it off.
+axum = [
+    "dep:axum",
+    "dep:jsonwebtoken",
+    "dep:reqwest",
+    "dep:tokio",
+    "dep:tracing",
+]
 
 [dependencies]
 serde = { workspace = true }
@@ -37,3 +46,7 @@ opentelemetry-appender-tracing = { version = "0.29", optional = true }
 tracing-opentelemetry = { version = "0.30", optional = true }
 tracing-subscriber = { workspace = true, optional = true }
 tracing = { workspace = true, optional = true }
+axum = { version = "0.8", optional = true }
+jsonwebtoken = { version = "9", optional = true }
+reqwest = { workspace = true, optional = true }
+tokio = { workspace = true, optional = true }

compliance-core/src/auth.rs

@@ -0,0 +1,390 @@
+//! M7.1 — JWT validation + tenant context propagation.
+//!
+//! `require_jwt_auth` validates a Bearer JWT against Keycloak's JWKS and
+//! attaches a [`TenantContext`] to the request extensions. Downstream
+//! middleware ([`require_tenant_status`]) and Axum extractors
+//! ([`crate::tenant_ctx::TenantCtx`]) read it from there.
+//!
+//! Skipped paths:
+//! * `/api/v1/health` — Kubernetes liveness; never authenticated.
+//!
+//! Failure modes:
+//! * No `JwksState` extension → pass-through (single-tenant dev mode).
+//! * Missing / malformed Bearer header → 401.
+//! * Signature / expiry invalid → 401.
+//! * Claims present but tenant_id missing → 401 (treated as a malformed
+//!   token; the realm must always issue tenant_id).
+
+use std::sync::Arc;
+
+use axum::{
+    extract::Request,
+    http::Method,
+    middleware::Next,
+    response::{IntoResponse, Response},
+};
+use jsonwebtoken::{decode, decode_header, jwk::JwkSet, DecodingKey, Validation};
+use reqwest::StatusCode;
+use serde::Deserialize;
+use tokio::sync::RwLock;
+
+use crate::{OrgRole, TenantContext, TenantStatus};
+
+/// Cached JWKS from Keycloak for token validation.
+#[derive(Clone)]
+pub struct JwksState {
+    pub jwks: Arc<RwLock<Option<JwkSet>>>,
+    pub jwks_url: String,
+}
+
+/// Raw shape of the JWT payload — matches the breakpilot-dev realm's
+/// protocol-mapper output. Missing fields default to "" / empty so a
+/// realm that hasn't been fully wired yet still validates.
+#[derive(Debug, Deserialize)]
+struct Claims {
+    sub: String,
+    #[serde(default)]
+    name: Option<String>,
+    #[serde(default)]
+    preferred_username: Option<String>,
+    #[serde(default)]
+    tenant_id: String,
+    #[serde(default)]
+    tenant_slug: String,
+    #[serde(default)]
+    org_roles: Vec<String>,
+    #[serde(default)]
+    products: Vec<String>,
+    #[serde(default)]
+    plan: String,
+    #[serde(default)]
+    tenant_status: Option<TenantStatus>,
+}
+
+const PUBLIC_ENDPOINTS: &[&str] = &["/api/v1/health"];
+
+/// Middleware that validates Bearer JWT tokens against Keycloak's JWKS
+/// and attaches a `TenantContext` extension on success.
+///
+/// Skips validation for the health endpoint.
+/// If `JwksState` is not present (Keycloak not configured), requests
+/// pass through and downstream code must handle the missing context.
+pub async fn require_jwt_auth(mut request: Request, next: Next) -> Response {
+    let path = request.uri().path();
+
+    if PUBLIC_ENDPOINTS.contains(&path) {
+        return next.run(request).await;
+    }
+
+    let jwks_state = match request.extensions().get::<JwksState>() {
+        Some(s) => s.clone(),
+        None => return next.run(request).await,
+    };
+
+    let auth_header = match request.headers().get("authorization") {
+        Some(h) => h,
+        None => return (StatusCode::UNAUTHORIZED, "Missing authorization header").into_response(),
+    };
+
+    let token = match auth_header.to_str() {
+        Ok(s) if s.starts_with("Bearer ") => &s[7..],
+        _ => return (StatusCode::UNAUTHORIZED, "Invalid authorization header").into_response(),
+    };
+
+    match validate_token(token, &jwks_state).await {
+        Ok(ctx) => {
+            request.extensions_mut().insert(ctx);
+            next.run(request).await
+        }
+        Err(e) => {
+            tracing::warn!("JWT validation failed: {e}");
+            (StatusCode::UNAUTHORIZED, "Invalid token").into_response()
+        }
+    }
+}
+
+/// Middleware that enforces the M7.1 `tenant_status` contract.
+///
+/// * `Active` / `Trial` / `Demo` — pass through.
+/// * `Frozen` — read-only after cancel / non-payment. Writes return 402.
+/// * `Archived` — data-retention window closed. Every request returns 410.
+///
+/// Pass-through when no `TenantContext` is present (single-tenant dev or
+/// the upstream JWT middleware ran without `JwksState`).
+pub async fn require_tenant_status(request: Request, next: Next) -> Response {
+    let ctx = match request.extensions().get::<TenantContext>() {
+        Some(c) => c.clone(),
+        None => return next.run(request).await,
+    };
+
+    if ctx.status.is_archived() {
+        return (
+            StatusCode::GONE,
+            "Tenant archived — data retention window closed",
+        )
+            .into_response();
+    }
+
+    if ctx.status.is_frozen() && is_write(request.method()) {
+        return (
+            StatusCode::PAYMENT_REQUIRED,
+            "Tenant frozen — read-only. Re-activate to resume writes.",
+        )
+            .into_response();
+    }
+
+    next.run(request).await
+}
+
+/// Treat anything other than GET/HEAD/OPTIONS as a write. Good enough for
+/// REST. The few exceptions (e.g. read-side POSTs) can opt out at the
+/// handler level once we have them.
+fn is_write(m: &Method) -> bool {
+    !matches!(m, &Method::GET | &Method::HEAD | &Method::OPTIONS)
+}
+
+async fn validate_token(token: &str, state: &JwksState) -> Result<TenantContext, String> {
+    let header = decode_header(token).map_err(|e| format!("failed to decode JWT header: {e}"))?;
+
+    let kid = header
+        .kid
+        .clone()
+        .ok_or_else(|| "JWT missing kid header".to_string())?;
+
+    // First try against whatever's currently cached. If the kid isn't
+    // there or the signature doesn't verify, the cached JWKS is most
+    // likely stale (KC rotated keys) — refresh once and retry before
+    // giving up. Without this every key rotation produces a silent 401
+    // storm that only goes away when the agent restarts.
+    let jwks = fetch_or_get_jwks(state, false).await?;
+    match try_validate(token, &header, &kid, &jwks) {
+        Ok(ctx) => Ok(ctx),
+        Err(ValidationError::Permanent(e)) => Err(e),
+        Err(ValidationError::Stale(reason)) => {
+            tracing::info!(
+                kid = %kid,
+                reason = %reason,
+                "JWKS appears stale — forcing refresh and retrying"
+            );
+            let jwks = fetch_or_get_jwks(state, true).await?;
+            try_validate(token, &header, &kid, &jwks).map_err(|e| match e {
+                ValidationError::Stale(s) | ValidationError::Permanent(s) => s,
+            })
+        }
+    }
+}
+
+#[derive(Debug)]
+enum ValidationError {
+    /// Refresh-eligible: cached JWKS may be stale.
+    Stale(String),
+    /// Refusing the token regardless of JWKS freshness.
+    Permanent(String),
+}
+
+fn try_validate(
+    token: &str,
+    header: &jsonwebtoken::Header,
+    kid: &str,
+    jwks: &JwkSet,
+) -> Result<TenantContext, ValidationError> {
+    let jwk = match jwks
+        .keys
+        .iter()
+        .find(|k| k.common.key_id.as_deref() == Some(kid))
+    {
+        Some(j) => j,
+        None => {
+            return Err(ValidationError::Stale(
+                "no matching key found in JWKS".to_string(),
+            ))
+        }
+    };
+
+    let decoding_key = DecodingKey::from_jwk(jwk)
+        .map_err(|e| ValidationError::Permanent(format!("failed to create decoding key: {e}")))?;
+
+    let mut validation = Validation::new(header.alg);
+    validation.validate_exp = true;
+    validation.validate_aud = false;
+
+    let data = match decode::<Claims>(token, &decoding_key, &validation) {
+        Ok(d) => d,
+        Err(e) => {
+            // Signature mismatch is the other refresh-eligible failure:
+            // the matching kid is present but the key bytes don't match.
+            // Everything else (expired, malformed, etc.) is permanent.
+            return Err(
+                if matches!(e.kind(), jsonwebtoken::errors::ErrorKind::InvalidSignature) {
+                    ValidationError::Stale(format!("token validation failed: {e}"))
+                } else {
+                    ValidationError::Permanent(format!("token validation failed: {e}"))
+                },
+            );
+        }
+    };
+
+    claims_to_context(data.claims).map_err(ValidationError::Permanent)
+}
+
+/// Map the decoded JWT payload into the platform-wide `TenantContext`.
+/// Pulled out for unit testing — no I/O.
+fn claims_to_context(c: Claims) -> Result<TenantContext, String> {
+    if c.tenant_id.is_empty() {
+        return Err("JWT is missing tenant_id claim".to_string());
+    }
+
+    let status = c.tenant_status.unwrap_or_else(|| {
+        tracing::warn!(
+            "JWT missing tenant_status claim for tenant {} — defaulting to Trial",
+            c.tenant_id
+        );
+        TenantStatus::Trial
+    });
+
+    Ok(TenantContext {
+        tenant_id: c.tenant_id,
+        tenant_slug: c.tenant_slug,
+        org_roles: c.org_roles.iter().map(|r| OrgRole::parse(r)).collect(),
+        products: c.products,
+        plan: c.plan,
+        status,
+        user_id: c.sub,
+        user_name: c.name.or(c.preferred_username),
+    })
+}
+
+async fn fetch_or_get_jwks(state: &JwksState, force: bool) -> Result<JwkSet, String> {
+    if !force {
+        let cached = state.jwks.read().await;
+        if let Some(ref jwks) = *cached {
+            return Ok(jwks.clone());
+        }
+    }
+
+    // Hold the write lock across the fetch so concurrent refreshers
+    // don't all hammer Keycloak when keys rotate. If another writer
+    // already populated a fresh JWKS while we were waiting (and we
+    // weren't asked to force), use theirs.
+    let mut cached = state.jwks.write().await;
+    if !force {
+        if let Some(ref jwks) = *cached {
+            return Ok(jwks.clone());
+        }
+    }
+
+    let resp = reqwest::get(&state.jwks_url)
+        .await
+        .map_err(|e| format!("failed to fetch JWKS: {e}"))?;
+
+    let jwks: JwkSet = resp
+        .json()
+        .await
+        .map_err(|e| format!("failed to parse JWKS: {e}"))?;
+
+    *cached = Some(jwks.clone());
+
+    Ok(jwks)
+}
+
+#[cfg(test)]
+#[allow(clippy::expect_used, clippy::unwrap_used)]
+mod tests {
+    use super::*;
+
+    fn base_claims() -> Claims {
+        Claims {
+            sub: "user-123".to_string(),
+            name: Some("Alice Acme".to_string()),
+            preferred_username: None,
+            tenant_id: "00000000-0000-0000-0000-000000000001".to_string(),
+            tenant_slug: "acme".to_string(),
+            org_roles: vec!["IT_ADMIN".to_string()],
+            products: vec!["compliance".to_string()],
+            plan: "professional".to_string(),
+            tenant_status: Some(TenantStatus::Active),
+        }
+    }
+
+    #[test]
+    fn claims_to_context_happy_path() {
+        let ctx = claims_to_context(base_claims()).expect("should map");
+        assert_eq!(ctx.tenant_id, "00000000-0000-0000-0000-000000000001");
+        assert_eq!(ctx.tenant_slug, "acme");
+        assert_eq!(ctx.org_roles, vec![OrgRole::ItAdmin]);
+        assert_eq!(ctx.products, vec!["compliance"]);
+        assert_eq!(ctx.plan, "professional");
+        assert_eq!(ctx.status, TenantStatus::Active);
+        assert_eq!(ctx.user_id, "user-123");
+        assert_eq!(ctx.user_name.as_deref(), Some("Alice Acme"));
+    }
+
+    #[test]
+    fn claims_to_context_rejects_missing_tenant_id() {
+        let mut c = base_claims();
+        c.tenant_id = "".to_string();
+        let err = claims_to_context(c).expect_err("should reject");
+        assert!(err.contains("tenant_id"));
+    }
+
+    #[test]
+    fn claims_to_context_defaults_status_when_missing() {
+        let mut c = base_claims();
+        c.tenant_status = None;
+        let ctx = claims_to_context(c).expect("should map");
+        assert_eq!(ctx.status, TenantStatus::Trial);
+    }
+
+    #[test]
+    fn claims_to_context_falls_back_to_preferred_username() {
+        let mut c = base_claims();
+        c.name = None;
+        c.preferred_username = Some("alice@acme.dev".to_string());
+        let ctx = claims_to_context(c).expect("should map");
+        assert_eq!(ctx.user_name.as_deref(), Some("alice@acme.dev"));
+    }
+
+    #[test]
+    fn claims_to_context_parses_multiple_roles() {
+        let mut c = base_claims();
+        c.org_roles = vec![
+            "IT_ADMIN".to_string(),
+            "CXO".to_string(),
+            "GARBAGE".to_string(),
+        ];
+        let ctx = claims_to_context(c).expect("should map");
+        assert_eq!(
+            ctx.org_roles,
+            vec![OrgRole::ItAdmin, OrgRole::Cxo, OrgRole::Unknown]
+        );
+    }
+
+    #[test]
+    fn try_validate_returns_stale_when_kid_missing_from_jwks() {
+        // Empty JWKS — the kid we ask for can't possibly match. The error
+        // must classify as Stale so the caller refreshes JWKS and retries.
+        let jwks = JwkSet { keys: vec![] };
+        let header = jsonwebtoken::Header {
+            alg: jsonwebtoken::Algorithm::RS256,
+            kid: Some("kid-rotated-out".to_string()),
+            ..Default::default()
+        };
+        let err = try_validate("ignored.token.value", &header, "kid-rotated-out", &jwks)
+            .expect_err("should fail");
+        match err {
+            ValidationError::Stale(s) => assert!(s.contains("no matching key")),
+            ValidationError::Permanent(s) => panic!("must be Stale, got Permanent: {s}"),
+        }
+    }
+
+    #[test]
+    fn is_write_detects_methods() {
+        assert!(!is_write(&Method::GET));
+        assert!(!is_write(&Method::HEAD));
+        assert!(!is_write(&Method::OPTIONS));
+        assert!(is_write(&Method::POST));
+        assert!(is_write(&Method::PUT));
+        assert!(is_write(&Method::PATCH));
+        assert!(is_write(&Method::DELETE));
+    }
+}

compliance-core/src/config.rs

@@ -33,6 +33,8 @@ pub struct AgentConfig {
     pub pentest_verification_email: Option<String>,
     pub pentest_imap_host: Option<String>,
     pub pentest_imap_port: Option<u16>,
+    /// Use implicit TLS (IMAPS, port 993) instead of plain IMAP.
+    pub pentest_imap_tls: bool,
     pub pentest_imap_username: Option<String>,
     pub pentest_imap_password: Option<SecretString>,
 }

compliance-core/src/db.rs

@@ -0,0 +1,75 @@
+//! Database helpers shared across the workspace.
+//!
+//! `tenant_filter` returns the BSON filter that every query and update
+//! against a tenant-scoped collection MUST include. Centralising it here
+//! makes the rule grep-able and keeps query call-sites from accidentally
+//! omitting it.
+//!
+//! Future work (M7.2+): each collection model grows a `tenant_id` field
+//! and every `find` / `update_*` / `delete_*` call gets this filter
+//! merged in. The migration to per-collection scoping is tracked
+//! separately — this helper is the building block.
+
+use bson::{doc, Document};
+
+use crate::TenantContext;
+
+/// Returns `{ "tenant_id": <ctx.tenant_id> }`. Merge this into every
+/// query filter against a tenant-scoped collection.
+///
+/// Use [`tenant_filter_merge`] when you need to combine it with other
+/// query conditions — it preserves both halves without overwriting.
+pub fn tenant_filter(ctx: &TenantContext) -> Document {
+    doc! { "tenant_id": &ctx.tenant_id }
+}
+
+/// Returns the tenant filter merged with caller-supplied conditions.
+/// The tenant_id always wins on key conflict — callers cannot
+/// accidentally override the scoping.
+pub fn tenant_filter_merge(ctx: &TenantContext, mut extra: Document) -> Document {
+    extra.insert("tenant_id", &ctx.tenant_id);
+    extra
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::TenantStatus;
+
+    fn ctx() -> TenantContext {
+        TenantContext {
+            tenant_id: "t-abc".to_string(),
+            tenant_slug: "acme".to_string(),
+            org_roles: vec![],
+            products: vec![],
+            plan: "starter".to_string(),
+            status: TenantStatus::Active,
+            user_id: "u-1".to_string(),
+            user_name: None,
+        }
+    }
+
+    #[test]
+    fn produces_tenant_id_filter() {
+        let f = tenant_filter(&ctx());
+        assert_eq!(f.get_str("tenant_id"), Ok("t-abc"));
+        assert_eq!(f.len(), 1);
+    }
+
+    #[test]
+    fn merge_preserves_extra_conditions() {
+        let extra = doc! { "status": "open", "severity": "high" };
+        let f = tenant_filter_merge(&ctx(), extra);
+        assert_eq!(f.get_str("tenant_id"), Ok("t-abc"));
+        assert_eq!(f.get_str("status"), Ok("open"));
+        assert_eq!(f.get_str("severity"), Ok("high"));
+    }
+
+    #[test]
+    fn merge_overrides_caller_tenant_id() {
+        let extra = doc! { "tenant_id": "evil-other", "status": "open" };
+        let f = tenant_filter_merge(&ctx(), extra);
+        assert_eq!(f.get_str("tenant_id"), Ok("t-abc"));
+        assert_eq!(f.get_str("status"), Ok("open"));
+    }
+}

compliance-core/src/lib.rs

@@ -1,9 +1,17 @@
 pub mod config;
+pub mod db;
 pub mod error;
 pub mod models;
 #[cfg(feature = "telemetry")]
 pub mod telemetry;
+pub mod tenant;
 pub mod traits;
 
+#[cfg(feature = "axum")]
+pub mod auth;
+#[cfg(feature = "axum")]
+pub mod tenant_ctx;
+
 pub use config::{AgentConfig, DashboardConfig};
 pub use error::CoreError;
+pub use tenant::{OrgRole, TenantContext, TenantStatus};

compliance-core/src/models/mcp_token.rs

@@ -0,0 +1,69 @@
+//! Per-tenant API tokens used by `compliance-mcp` to authenticate MCP
+//! HTTP requests on behalf of LLM clients (Claude Desktop, Cursor,
+//! ChatGPT, etc.) that can't run a Keycloak OIDC flow.
+//!
+//! Tokens are opaque strings of the form `mcpt_<44 url-safe random
+//! chars>`. The raw value is shown to the user exactly once at
+//! creation; the database only ever sees the SHA-256 hash. Lookups go
+//! through the cross-tenant `<prefix>__admin.mcp_tokens` collection
+//! and return the `tenant_id` the MCP server should route to.
+
+use chrono::{DateTime, Utc};
+use serde::{Deserialize, Serialize};
+
+/// Persisted token metadata. `token_hash` is the SHA-256 hex of the
+/// raw token; the raw token itself is never stored.
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct McpToken {
+    #[serde(rename = "_id", skip_serializing_if = "Option::is_none")]
+    pub id: Option<bson::oid::ObjectId>,
+    /// SHA-256 hex of the raw token. Unique index in the collection.
+    pub token_hash: String,
+    /// First 8 chars of the raw token — purely for UI display so users
+    /// can identify which token is which without re-issuing.
+    pub token_prefix: String,
+    /// Routes to `<db_prefix>_<tenant_id>` on MCP requests.
+    pub tenant_id: String,
+    /// User-given label, e.g. "Claude Desktop" or "Sharang's laptop".
+    pub name: String,
+    /// Keycloak `sub` of the user who created this token, for audit.
+    pub created_by: String,
+    #[serde(with = "super::serde_helpers::bson_datetime")]
+    pub created_at: DateTime<Utc>,
+    #[serde(default, with = "super::serde_helpers::opt_bson_datetime")]
+    pub last_used_at: Option<DateTime<Utc>>,
+    /// Soft-delete flag. A revoked token doc stays around for audit
+    /// but never authenticates.
+    #[serde(default)]
+    pub revoked: bool,
+}
+
+/// Public projection of a token — never includes the hash.
+/// Returned by `GET /api/v1/mcp-tokens`.
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct McpTokenView {
+    pub id: String,
+    pub name: String,
+    /// `mcpt_xxxx…` so the user can identify which row is which.
+    pub token_prefix: String,
+    pub created_by: String,
+    #[serde(with = "super::serde_helpers::bson_datetime")]
+    pub created_at: DateTime<Utc>,
+    #[serde(default, with = "super::serde_helpers::opt_bson_datetime")]
+    pub last_used_at: Option<DateTime<Utc>>,
+    pub revoked: bool,
+}
+
+impl From<&McpToken> for McpTokenView {
+    fn from(t: &McpToken) -> Self {
+        Self {
+            id: t.id.map(|o| o.to_hex()).unwrap_or_default(),
+            name: t.name.clone(),
+            token_prefix: t.token_prefix.clone(),
+            created_by: t.created_by.clone(),
+            created_at: t.created_at,
+            last_used_at: t.last_used_at,
+            revoked: t.revoked,
+        }
+    }
+}

compliance-core/src/models/mod.rs

@@ -7,6 +7,8 @@ pub mod finding;
 pub mod graph;
 pub mod issue;
 pub mod mcp;
+pub mod mcp_token;
+pub mod notification;
 pub mod pentest;
 pub mod repository;
 pub mod sbom;
@@ -27,6 +29,8 @@ pub use graph::{
 };
 pub use issue::{IssueStatus, TrackerIssue, TrackerType};
 pub use mcp::{McpServerConfig, McpServerStatus, McpTransport};
+pub use mcp_token::{McpToken, McpTokenView};
+pub use notification::{CveNotification, NotificationSeverity, NotificationStatus};
 pub use pentest::{
     AttackChainNode, AttackNodeStatus, AuthMode, CodeContextHint, Environment, IdentityProvider,
     PentestAuthConfig, PentestConfig, PentestEvent, PentestMessage, PentestSession, PentestStats,

compliance-core/src/models/notification.rs

@@ -0,0 +1,103 @@
+use chrono::{DateTime, Utc};
+use serde::{Deserialize, Serialize};
+
+/// Status of a CVE notification
+#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
+#[serde(rename_all = "lowercase")]
+pub enum NotificationStatus {
+    /// Newly created, not yet seen by the user
+    New,
+    /// User has seen it (e.g., opened the notification panel)
+    Read,
+    /// User has explicitly acknowledged/dismissed it
+    Dismissed,
+}
+
+/// Severity level for notification filtering
+#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq, PartialOrd, Ord)]
+#[serde(rename_all = "lowercase")]
+pub enum NotificationSeverity {
+    Low,
+    Medium,
+    High,
+    Critical,
+}
+
+/// A notification about a newly discovered CVE affecting a tracked dependency.
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct CveNotification {
+    #[serde(rename = "_id", skip_serializing_if = "Option::is_none")]
+    pub id: Option<bson::oid::ObjectId>,
+    /// The CVE/GHSA identifier
+    pub cve_id: String,
+    /// Repository where the vulnerable dependency is used
+    pub repo_id: String,
+    /// Repository name (denormalized for display)
+    pub repo_name: String,
+    /// Affected package name
+    pub package_name: String,
+    /// Affected version
+    pub package_version: String,
+    /// Human-readable severity
+    pub severity: NotificationSeverity,
+    /// CVSS score if available
+    pub cvss_score: Option<f64>,
+    /// Short summary of the vulnerability
+    pub summary: Option<String>,
+    /// Link to vulnerability details
+    pub url: Option<String>,
+    /// Notification lifecycle status
+    pub status: NotificationStatus,
+    /// When the CVE was first detected for this dependency
+    #[serde(with = "super::serde_helpers::bson_datetime")]
+    pub created_at: DateTime<Utc>,
+    /// When the user last interacted with this notification
+    pub read_at: Option<DateTime<Utc>>,
+}
+
+impl CveNotification {
+    pub fn new(
+        cve_id: String,
+        repo_id: String,
+        repo_name: String,
+        package_name: String,
+        package_version: String,
+        severity: NotificationSeverity,
+    ) -> Self {
+        Self {
+            id: None,
+            cve_id,
+            repo_id,
+            repo_name,
+            package_name,
+            package_version,
+            severity,
+            cvss_score: None,
+            summary: None,
+            url: None,
+            status: NotificationStatus::New,
+            created_at: Utc::now(),
+            read_at: None,
+        }
+    }
+}
+
+/// Map an OSV/NVD severity string to our notification severity
+pub fn parse_severity(s: Option<&str>, cvss: Option<f64>) -> NotificationSeverity {
+    // Prefer CVSS score if available
+    if let Some(score) = cvss {
+        return match score {
+            s if s >= 9.0 => NotificationSeverity::Critical,
+            s if s >= 7.0 => NotificationSeverity::High,
+            s if s >= 4.0 => NotificationSeverity::Medium,
+            _ => NotificationSeverity::Low,
+        };
+    }
+    // Fall back to string severity
+    match s.map(|s| s.to_uppercase()).as_deref() {
+        Some("CRITICAL") => NotificationSeverity::Critical,
+        Some("HIGH") => NotificationSeverity::High,
+        Some("MODERATE" | "MEDIUM") => NotificationSeverity::Medium,
+        _ => NotificationSeverity::Low,
+    }
+}

compliance-core/src/tenant.rs

@@ -0,0 +1,165 @@
+//! Tenant context propagated through every authenticated request.
+//!
+//! M7.1 single source of truth for "who is this request for". Claims come
+//! from a Keycloak-issued JWT and land here via [`crate::auth::require_jwt_auth`]
+//! (enabled with the `axum` feature). Handlers reach into the request
+//! extensions with the [`crate::tenant_ctx::TenantCtx`] extractor.
+//!
+//! The shape mirrors the JWT claim names the breakpilot-platform realm
+//! emits (see `platform/orca-platform/dev/keycloak/realm-export.json`).
+//! Stable contract — adding fields is fine; renaming is a breaking
+//! change for every downstream product.
+
+use serde::{Deserialize, Serialize};
+
+/// Tenant lifecycle status from `PLATFORM_ARCHITECTURE.md §5c`.
+///
+/// Drives the `tenant_status` middleware:
+/// * `Demo` / `Trial` / `Active` — full access.
+/// * `Frozen` — read-only after cancel / non-payment. Mutating endpoints
+///   return 402.
+/// * `Archived` — data-retention window closed. Every endpoint returns 410.
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(rename_all = "lowercase")]
+pub enum TenantStatus {
+    Demo,
+    Trial,
+    Active,
+    Frozen,
+    Archived,
+}
+
+impl TenantStatus {
+    /// True for statuses that block write paths.
+    pub fn is_frozen(&self) -> bool {
+        matches!(self, TenantStatus::Frozen)
+    }
+    /// True for statuses that block every request.
+    pub fn is_archived(&self) -> bool {
+        matches!(self, TenantStatus::Archived)
+    }
+    /// True for the shared demo tenant — metering, billing, and audit
+    /// export are skipped.
+    pub fn is_demo(&self) -> bool {
+        matches!(self, TenantStatus::Demo)
+    }
+}
+
+impl std::fmt::Display for TenantStatus {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            Self::Demo => write!(f, "demo"),
+            Self::Trial => write!(f, "trial"),
+            Self::Active => write!(f, "active"),
+            Self::Frozen => write!(f, "frozen"),
+            Self::Archived => write!(f, "archived"),
+        }
+    }
+}
+
+/// Org-level role baked into the JWT by the realm's protocol mapper.
+/// `PLATFORM_ARCHITECTURE.md §6` is the canonical list.
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(rename_all = "UPPERCASE")]
+pub enum OrgRole {
+    ItAdmin,
+    Cxo,
+    Finance,
+    Legal,
+    User,
+    /// Anything we haven't enumerated yet — forwards-compatible.
+    #[serde(other)]
+    Unknown,
+}
+
+impl OrgRole {
+    /// Parses a single role string (Keycloak emits these as `IT_ADMIN`,
+    /// `CXO`, etc.). Round-trips with the JSON layer.
+    pub fn parse(s: &str) -> Self {
+        match s {
+            "IT_ADMIN" => OrgRole::ItAdmin,
+            "CXO" => OrgRole::Cxo,
+            "FINANCE" => OrgRole::Finance,
+            "LEGAL" => OrgRole::Legal,
+            "USER" => OrgRole::User,
+            _ => OrgRole::Unknown,
+        }
+    }
+}
+
+/// Everything we know about the requesting tenant at the moment a request
+/// lands. Cheap to clone (every field is owned + small).
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct TenantContext {
+    /// `tenants.id` from the platform's tenant-registry (UUID).
+    pub tenant_id: String,
+    /// Lowercase URL-safe slug. Useful for log lines + audit emit.
+    pub tenant_slug: String,
+    /// Org-level roles the authenticated user holds inside this tenant.
+    /// Drives the per-handler RBAC in `M7.1-followup` PRs.
+    pub org_roles: Vec<OrgRole>,
+    /// Products this tenant is currently entitled to. Used to short-circuit
+    /// MCP / API calls for unsubscribed products.
+    pub products: Vec<String>,
+    /// Customer plan (`starter` / `professional` / `enterprise`) — gates
+    /// per-plan feature flags (e.g., MCP server is enterprise-only).
+    pub plan: String,
+    /// Lifecycle status — read by `require_tenant_status` middleware.
+    pub status: TenantStatus,
+    /// Keycloak user id of the requester (`sub` claim). Required for audit
+    /// emit so we know WHO did the thing, not just WHICH tenant.
+    pub user_id: String,
+    /// Optional user-facing name from the `name` / `preferred_username`
+    /// claim. Only used in audit + log lines.
+    pub user_name: Option<String>,
+}
+
+impl TenantContext {
+    /// True if the caller holds at least one of the listed roles.
+    pub fn has_any_role(&self, roles: &[OrgRole]) -> bool {
+        self.org_roles.iter().any(|r| roles.contains(r))
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn org_role_parses_known_values() {
+        assert_eq!(OrgRole::parse("IT_ADMIN"), OrgRole::ItAdmin);
+        assert_eq!(OrgRole::parse("CXO"), OrgRole::Cxo);
+        assert_eq!(OrgRole::parse("USER"), OrgRole::User);
+    }
+
+    #[test]
+    fn org_role_unknown_is_forward_compat() {
+        assert_eq!(OrgRole::parse("FUTURE_ROLE"), OrgRole::Unknown);
+    }
+
+    #[test]
+    fn tenant_status_predicates() {
+        assert!(TenantStatus::Frozen.is_frozen());
+        assert!(!TenantStatus::Active.is_frozen());
+        assert!(TenantStatus::Archived.is_archived());
+        assert!(TenantStatus::Demo.is_demo());
+        assert!(!TenantStatus::Active.is_demo());
+    }
+
+    #[test]
+    fn has_any_role_matches() {
+        let ctx = TenantContext {
+            tenant_id: "t1".into(),
+            tenant_slug: "acme".into(),
+            org_roles: vec![OrgRole::ItAdmin],
+            products: vec![],
+            plan: "professional".into(),
+            status: TenantStatus::Active,
+            user_id: "u".into(),
+            user_name: None,
+        };
+        assert!(ctx.has_any_role(&[OrgRole::ItAdmin]));
+        assert!(ctx.has_any_role(&[OrgRole::Cxo, OrgRole::ItAdmin]));
+        assert!(!ctx.has_any_role(&[OrgRole::User, OrgRole::Cxo]));
+    }
+}

compliance-core/src/tenant_ctx.rs

@@ -0,0 +1,95 @@
+//! Axum extractor for the per-request `TenantContext`.
+//!
+//! Handlers consume it as a normal extractor argument:
+//!
+//! ```ignore
+//! async fn list_findings(TenantCtx(ctx): TenantCtx) -> Json<...> {
+//!     let filter = compliance_core::db::tenant_filter(&ctx);
+//!     ...
+//! }
+//! ```
+//!
+//! The middleware ([`crate::auth::require_jwt_auth`]) is responsible for
+//! inserting the context into the request extensions. If it's missing on
+//! a route that uses this extractor, that's a bug in the wiring — we
+//! return 401 so the caller sees an auth failure rather than a 500.
+
+use axum::{
+    extract::FromRequestParts,
+    http::{request::Parts, StatusCode},
+    response::{IntoResponse, Response},
+};
+
+use crate::TenantContext;
+
+#[derive(Debug, Clone)]
+pub struct TenantCtx(pub TenantContext);
+
+#[derive(Debug)]
+pub struct TenantCtxRejection;
+
+impl IntoResponse for TenantCtxRejection {
+    fn into_response(self) -> Response {
+        (
+            StatusCode::UNAUTHORIZED,
+            "Missing tenant context — request was not authenticated",
+        )
+            .into_response()
+    }
+}
+
+impl<S> FromRequestParts<S> for TenantCtx
+where
+    S: Send + Sync,
+{
+    type Rejection = TenantCtxRejection;
+
+    async fn from_request_parts(parts: &mut Parts, _state: &S) -> Result<Self, Self::Rejection> {
+        parts
+            .extensions
+            .get::<TenantContext>()
+            .cloned()
+            .map(TenantCtx)
+            .ok_or(TenantCtxRejection)
+    }
+}
+
+#[cfg(test)]
+#[allow(clippy::expect_used, clippy::unwrap_used)]
+mod tests {
+    use super::*;
+    use crate::TenantStatus;
+    use axum::http::Request;
+
+    fn ctx() -> TenantContext {
+        TenantContext {
+            tenant_id: "t-1".to_string(),
+            tenant_slug: "acme".to_string(),
+            org_roles: vec![],
+            products: vec![],
+            plan: "starter".to_string(),
+            status: TenantStatus::Active,
+            user_id: "u-1".to_string(),
+            user_name: None,
+        }
+    }
+
+    #[tokio::test]
+    async fn extracts_context_when_present() {
+        let mut req = Request::new(());
+        req.extensions_mut().insert(ctx());
+        let (mut parts, _) = req.into_parts();
+        let TenantCtx(found) = TenantCtx::from_request_parts(&mut parts, &())
+            .await
+            .expect("extractor should succeed");
+        assert_eq!(found.tenant_id, "t-1");
+    }
+
+    #[tokio::test]
+    async fn rejects_when_missing() {
+        let req: Request<()> = Request::new(());
+        let (mut parts, _) = req.into_parts();
+        let err = TenantCtx::from_request_parts(&mut parts, &()).await;
+        assert!(err.is_err());
+    }
+}

compliance-dashboard/Cargo.toml

@@ -51,7 +51,7 @@ thiserror = { workspace = true }
 
 # Web-only
 reqwest = { workspace = true, optional = true }
-web-sys = { version = "0.3", optional = true, features = ["Blob", "BlobPropertyBag", "HtmlAnchorElement", "Url", "Document", "Window"] }
+web-sys = { version = "0.3", optional = true, features = ["Blob", "BlobPropertyBag", "HtmlAnchorElement", "Url", "Document", "Element", "Window", "Storage", "MediaQueryList"] }
 js-sys = { version = "0.3", optional = true }
 wasm-bindgen = { version = "0.2", optional = true }
 gloo-timers = { version = "0.3", features = ["futures"], optional = true }

compliance-dashboard/assets/main.css

@@ -61,6 +61,77 @@
     --ease-spring: cubic-bezier(0.34, 1.56, 0.64, 1);
 }
 
+/* ── Light theme tokens ──
+   Applied when the user has explicitly chosen light (`data-theme="light"`)
+   OR when their OS prefers light AND they have made no explicit choice. */
+:root[data-theme="light"] {
+    --bg-primary: #f5f7fb;
+    --bg-secondary: #ffffff;
+    --bg-card: rgba(255, 255, 255, 0.85);
+    --bg-card-solid: #ffffff;
+    --bg-card-hover: #f1f5fb;
+    --bg-elevated: #f8fafc;
+
+    --text-primary: #0c1426;
+    --text-secondary: #475569;
+    --text-tertiary: #8a9bb4;
+
+    --accent: #0070d4;
+    --accent-hover: #0080f0;
+    --accent-muted: rgba(0, 112, 212, 0.10);
+    --accent-glow: 0 0 20px rgba(0, 112, 212, 0.10);
+
+    --border: #e2e8f0;
+    --border-bright: #cbd5e1;
+    --border-accent: rgba(0, 112, 212, 0.30);
+
+    --danger: #dc2626;
+    --danger-bg: rgba(220, 38, 38, 0.08);
+    --warning: #d97706;
+    --warning-bg: rgba(217, 119, 6, 0.08);
+    --success: #16a34a;
+    --success-bg: rgba(22, 163, 74, 0.08);
+    --info: #2563eb;
+    --info-bg: rgba(37, 99, 235, 0.08);
+    --orange: #ea580c;
+    --orange-bg: rgba(234, 88, 12, 0.08);
+}
+
+@media (prefers-color-scheme: light) {
+    :root:not([data-theme="dark"]) {
+        --bg-primary: #f5f7fb;
+        --bg-secondary: #ffffff;
+        --bg-card: rgba(255, 255, 255, 0.85);
+        --bg-card-solid: #ffffff;
+        --bg-card-hover: #f1f5fb;
+        --bg-elevated: #f8fafc;
+
+        --text-primary: #0c1426;
+        --text-secondary: #475569;
+        --text-tertiary: #8a9bb4;
+
+        --accent: #0070d4;
+        --accent-hover: #0080f0;
+        --accent-muted: rgba(0, 112, 212, 0.10);
+        --accent-glow: 0 0 20px rgba(0, 112, 212, 0.10);
+
+        --border: #e2e8f0;
+        --border-bright: #cbd5e1;
+        --border-accent: rgba(0, 112, 212, 0.30);
+
+        --danger: #dc2626;
+        --danger-bg: rgba(220, 38, 38, 0.08);
+        --warning: #d97706;
+        --warning-bg: rgba(217, 119, 6, 0.08);
+        --success: #16a34a;
+        --success-bg: rgba(22, 163, 74, 0.08);
+        --info: #2563eb;
+        --info-bg: rgba(37, 99, 235, 0.08);
+        --orange: #ea580c;
+        --orange-bg: rgba(234, 88, 12, 0.08);
+    }
+}
+
 
 /* ── Reset & Base ── */
 
@@ -396,6 +467,44 @@ code {
     background: rgba(0, 200, 255, 0.06);
 }
 
+.theme-toggle {
+    background: none;
+    border: none;
+    border-top: 1px solid var(--border);
+    color: var(--text-secondary);
+    padding: 11px 18px;
+    cursor: pointer;
+    display: flex;
+    align-items: center;
+    gap: 11px;
+    font-family: var(--font-body);
+    font-size: 13.5px;
+    font-weight: 500;
+    transition: color 0.2s, background 0.2s;
+    width: 100%;
+    text-align: left;
+}
+
+.theme-toggle:hover {
+    color: var(--accent);
+    background: var(--accent-muted);
+}
+
+.theme-toggle svg {
+    flex-shrink: 0;
+    opacity: 0.75;
+    transition: opacity 0.2s;
+}
+
+.theme-toggle:hover svg {
+    opacity: 1;
+}
+
+.sidebar.collapsed .theme-toggle {
+    justify-content: center;
+    padding: 11px 0;
+}
+
 .sidebar.collapsed .sidebar-header {
     padding: 22px 0;
     justify-content: center;
@@ -3645,3 +3754,277 @@ tbody tr:last-child td {
 .wizard-toggle.active .wizard-toggle-knob {
     transform: translateX(16px);
 }
+
+/* ═══════════════════════════════════════════════════════════════
+   HELP CHAT WIDGET
+   Floating assistant for documentation Q&A
+   ═══════════════════════════════════════════════════════════════ */
+
+.help-chat-toggle {
+    position: fixed;
+    bottom: 24px;
+    right: 28px;
+    z-index: 50;
+    width: 48px;
+    height: 48px;
+    border-radius: 50%;
+    background: var(--accent);
+    color: var(--bg-primary);
+    border: none;
+    cursor: pointer;
+    display: flex;
+    align-items: center;
+    justify-content: center;
+    box-shadow: 0 4px 20px rgba(0, 200, 255, 0.3);
+    transition: transform 0.15s, box-shadow 0.15s;
+}
+.help-chat-toggle:hover {
+    transform: scale(1.08);
+    box-shadow: 0 6px 28px rgba(0, 200, 255, 0.4);
+}
+
+.help-chat-panel {
+    position: fixed;
+    bottom: 24px;
+    right: 28px;
+    z-index: 51;
+    width: 400px;
+    height: 520px;
+    background: var(--bg-secondary);
+    border: 1px solid var(--border-bright);
+    border-radius: 16px;
+    display: flex;
+    flex-direction: column;
+    overflow: hidden;
+    box-shadow: 0 12px 48px rgba(0, 0, 0, 0.5), var(--accent-glow);
+}
+
+.help-chat-header {
+    display: flex;
+    align-items: center;
+    justify-content: space-between;
+    padding: 14px 18px;
+    border-bottom: 1px solid var(--border);
+    background: var(--bg-primary);
+}
+.help-chat-title {
+    display: flex;
+    align-items: center;
+    gap: 8px;
+    font-family: 'Outfit', sans-serif;
+    font-weight: 600;
+    font-size: 14px;
+    color: var(--text-primary);
+}
+.help-chat-close {
+    background: none;
+    border: none;
+    color: var(--text-secondary);
+    cursor: pointer;
+    padding: 4px;
+    border-radius: 6px;
+    display: flex;
+}
+.help-chat-close:hover {
+    color: var(--text-primary);
+    background: var(--bg-elevated);
+}
+
+.help-chat-messages {
+    flex: 1;
+    overflow-y: auto;
+    padding: 16px;
+    display: flex;
+    flex-direction: column;
+    gap: 12px;
+}
+
+.help-chat-empty {
+    display: flex;
+    flex-direction: column;
+    align-items: center;
+    justify-content: center;
+    height: 100%;
+    text-align: center;
+    color: var(--text-secondary);
+    font-size: 13px;
+    gap: 8px;
+}
+.help-chat-hint {
+    font-size: 12px;
+    color: var(--text-tertiary);
+    font-style: italic;
+}
+
+.help-msg {
+    max-width: 88%;
+    animation: helpMsgIn 0.15s ease-out;
+}
+@keyframes helpMsgIn {
+    from { opacity: 0; transform: translateY(6px); }
+    to { opacity: 1; transform: translateY(0); }
+}
+.help-msg-user {
+    align-self: flex-end;
+}
+.help-msg-assistant {
+    align-self: flex-start;
+}
+.help-msg-content {
+    padding: 10px 14px;
+    border-radius: 12px;
+    font-size: 13px;
+    line-height: 1.55;
+    word-wrap: break-word;
+}
+.help-msg-user .help-msg-content {
+    background: var(--accent);
+    color: var(--bg-primary);
+    border-bottom-right-radius: 4px;
+}
+.help-msg-assistant .help-msg-content {
+    background: var(--bg-elevated);
+    color: var(--text-primary);
+    border: 1px solid var(--border);
+    border-bottom-left-radius: 4px;
+}
+.help-msg-assistant .help-msg-content code {
+    background: rgba(0, 200, 255, 0.1);
+    padding: 1px 5px;
+    border-radius: 3px;
+    font-family: 'JetBrains Mono', monospace;
+    font-size: 12px;
+}
+.help-msg-loading {
+    padding: 10px 14px;
+    border-radius: 12px;
+    background: var(--bg-elevated);
+    border: 1px solid var(--border);
+    border-bottom-left-radius: 4px;
+    color: var(--text-secondary);
+    font-size: 13px;
+    animation: helpPulse 1.2s ease-in-out infinite;
+}
+@keyframes helpPulse {
+    0%, 100% { opacity: 0.6; }
+    50% { opacity: 1; }
+}
+
+.help-chat-input {
+    display: flex;
+    align-items: center;
+    gap: 8px;
+    padding: 12px 14px;
+    border-top: 1px solid var(--border);
+    background: var(--bg-primary);
+}
+.help-chat-input input {
+    flex: 1;
+    background: var(--bg-elevated);
+    border: 1px solid var(--border);
+    border-radius: 8px;
+    padding: 10px 14px;
+    color: var(--text-primary);
+    font-size: 13px;
+    font-family: 'DM Sans', sans-serif;
+    outline: none;
+    transition: border-color 0.15s;
+}
+.help-chat-input input:focus {
+    border-color: var(--accent);
+}
+.help-chat-input input::placeholder {
+    color: var(--text-tertiary);
+}
+.help-chat-send {
+    width: 36px;
+    height: 36px;
+    border-radius: 8px;
+    background: var(--accent);
+    color: var(--bg-primary);
+    border: none;
+    cursor: pointer;
+    display: flex;
+    align-items: center;
+    justify-content: center;
+    transition: opacity 0.15s;
+}
+.help-chat-send:disabled {
+    opacity: 0.4;
+    cursor: not-allowed;
+}
+.help-chat-send:not(:disabled):hover {
+    background: var(--accent-hover);
+}
+
+/* ═══════════════════════════════════════════════════════════════
+   NOTIFICATION BELL — CVE alert dropdown
+   ═══════════════════════════════════════════════════════════════ */
+.notification-bell-wrapper { position: fixed; top: 16px; right: 28px; z-index: 48; }
+.notification-bell-btn { position: relative; background: var(--bg-elevated); border: 1px solid var(--border); border-radius: 10px; padding: 8px 10px; color: var(--text-secondary); cursor: pointer; display: flex; align-items: center; transition: color 0.15s, border-color 0.15s; }
+.notification-bell-btn:hover { color: var(--text-primary); border-color: var(--border-bright); }
+.notification-badge { position: absolute; top: -4px; right: -4px; background: var(--danger); color: #fff; font-size: 10px; font-weight: 700; min-width: 18px; height: 18px; border-radius: 9px; display: flex; align-items: center; justify-content: center; padding: 0 4px; font-family: 'Outfit', sans-serif; }
+.notification-panel { position: absolute; top: 44px; right: 0; width: 380px; max-height: 480px; background: var(--bg-secondary); border: 1px solid var(--border-bright); border-radius: 12px; overflow: hidden; box-shadow: 0 12px 48px rgba(0,0,0,0.5); display: flex; flex-direction: column; }
+.notification-panel-header { display: flex; align-items: center; justify-content: space-between; padding: 12px 16px; border-bottom: 1px solid var(--border); font-family: 'Outfit', sans-serif; font-weight: 600; font-size: 14px; color: var(--text-primary); }
+.notification-close-btn { background: none; border: none; color: var(--text-secondary); cursor: pointer; padding: 2px; }
+.notification-panel-body { overflow-y: auto; flex: 1; padding: 8px; }
+.notification-loading, .notification-empty { display: flex; flex-direction: column; align-items: center; justify-content: center; padding: 32px 16px; color: var(--text-secondary); font-size: 13px; gap: 8px; }
+.notification-item { padding: 10px 12px; border-radius: 8px; margin-bottom: 4px; background: var(--bg-card); border: 1px solid var(--border); transition: border-color 0.15s; }
+.notification-item:hover { border-color: var(--border-bright); }
+.notification-item-header { display: flex; align-items: center; gap: 8px; margin-bottom: 4px; }
+.notification-sev { font-size: 10px; font-weight: 700; padding: 2px 6px; border-radius: 4px; text-transform: uppercase; letter-spacing: 0.5px; font-family: 'Outfit', sans-serif; }
+.notification-sev.sev-critical { background: var(--danger-bg); color: var(--danger); }
+.notification-sev.sev-high { background: rgba(255,140,0,0.12); color: #ff8c00; }
+.notification-sev.sev-medium { background: var(--warning-bg); color: var(--warning); }
+.notification-sev.sev-low { background: rgba(0,200,255,0.08); color: var(--accent); }
+.notification-cve-id { font-size: 12px; font-weight: 600; color: var(--text-primary); font-family: 'JetBrains Mono', monospace; }
+.notification-cve-id a { color: var(--accent); text-decoration: none; }
+.notification-cve-id a:hover { text-decoration: underline; }
+.notification-cvss { font-size: 10px; color: var(--text-secondary); margin-left: auto; font-family: 'JetBrains Mono', monospace; }
+.notification-dismiss-btn { background: none; border: none; color: var(--text-tertiary); cursor: pointer; padding: 2px; margin-left: 4px; }
+.notification-dismiss-btn:hover { color: var(--danger); }
+.notification-item-pkg { font-size: 12px; color: var(--text-primary); font-family: 'JetBrains Mono', monospace; }
+.notification-item-repo { font-size: 11px; color: var(--text-secondary); margin-bottom: 4px; }
+.notification-item-summary { font-size: 11px; color: var(--text-secondary); line-height: 1.4; display: -webkit-box; -webkit-line-clamp: 2; -webkit-box-orient: vertical; overflow: hidden; }
+
+/* ═══════════════════════════════════════════════════════════════
+   COPY BUTTON — Reusable clipboard copy component
+   ═══════════════════════════════════════════════════════════════ */
+.copy-btn { background: none; border: 1px solid var(--border); border-radius: 6px; padding: 5px 7px; color: var(--text-secondary); cursor: pointer; display: inline-flex; align-items: center; transition: color 0.15s, border-color 0.15s, background 0.15s; flex-shrink: 0; }
+.copy-btn:hover { color: var(--accent); border-color: var(--accent); background: var(--accent-muted); }
+.copy-btn-sm { padding: 3px 5px; border-radius: 4px; }
+/* Copyable inline field pattern: value + copy button side by side */
+.copyable { display: flex; align-items: center; gap: 6px; }
+.copyable code, .copyable .mono { flex: 1; min-width: 0; overflow: hidden; text-overflow: ellipsis; white-space: nowrap; }
+.code-snippet-wrapper { position: relative; }
+.code-snippet-header { display: flex; align-items: center; justify-content: space-between; margin-bottom: 4px; gap: 8px; }
+
+
+/* ═══════════════════════════════════════════════════════════════
+   LIGHT THEME — surface overrides for the few hardcoded dark
+   colors that don't go through CSS custom properties.
+   ═══════════════════════════════════════════════════════════════ */
+
+:root[data-theme="light"] .main-content {
+    background-image: radial-gradient(circle at 1px 1px, rgba(100, 116, 139, 0.18) 1px, transparent 0);
+}
+:root[data-theme="light"] .code-block {
+    background: #f8fafc;
+    color: #0c1426;
+}
+:root[data-theme="light"] .graph-stab-overlay {
+    background: radial-gradient(ellipse at center, rgba(245, 247, 251, 0.92) 0%, rgba(245, 247, 251, 0.98) 100%);
+}
+
+@media (prefers-color-scheme: light) {
+    :root:not([data-theme="dark"]) .main-content {
+        background-image: radial-gradient(circle at 1px 1px, rgba(100, 116, 139, 0.18) 1px, transparent 0);
+    }
+    :root:not([data-theme="dark"]) .code-block {
+        background: #f8fafc;
+        color: #0c1426;
+    }
+    :root:not([data-theme="dark"]) .graph-stab-overlay {
+        background: radial-gradient(ellipse at center, rgba(245, 247, 251, 0.92) 0%, rgba(245, 247, 251, 0.98) 100%);
+    }
+}

compliance-dashboard/src/app.rs

@@ -44,8 +44,6 @@ pub enum Route {
         PentestSessionPage { session_id: String },
         #[route("/mcp-servers")]
         McpServersPage {},
-        #[route("/settings")]
-        SettingsPage {},
 }
 
 const FAVICON: Asset = asset!("/assets/favicon.svg");

compliance-dashboard/src/components/app_shell.rs

@@ -1,6 +1,8 @@
 use dioxus::prelude::*;
 
 use crate::app::Route;
+use crate::components::help_chat::HelpChat;
+use crate::components::notification_bell::NotificationBell;
 use crate::components::sidebar::Sidebar;
 use crate::components::toast::{ToastContainer, Toasts};
 use crate::infrastructure::auth_check::check_auth;
@@ -20,7 +22,9 @@ pub fn AppShell() -> Element {
                     main { class: "main-content",
                         Outlet::<Route> {}
                     }
+                    NotificationBell {}
                     ToastContainer {}
+                    HelpChat {}
                 }
             }
         }
@@ -28,7 +32,7 @@ pub fn AppShell() -> Element {
             // Not authenticated — redirect to Keycloak login
             rsx! {
                 document::Script {
-                    dangerous_inner_html: "window.location.href = '/auth';"
+                    "window.location.href = '/auth';"
                 }
             }
         }

compliance-dashboard/src/components/code_snippet.rs

@@ -1,5 +1,7 @@
 use dioxus::prelude::*;
 
+use crate::components::copy_button::CopyButton;
+
 #[component]
 pub fn CodeSnippet(
     code: String,
@@ -7,15 +9,18 @@ pub fn CodeSnippet(
     #[props(default)] line_number: u32,
 ) -> Element {
     rsx! {
-        div {
-            if !file_path.is_empty() {
-                div {
-                    style: "font-size: 12px; color: var(--text-secondary); margin-bottom: 4px; font-family: monospace;",
-                    "{file_path}"
-                    if line_number > 0 {
-                        ":{line_number}"
+        div { class: "code-snippet-wrapper",
+            div { class: "code-snippet-header",
+                if !file_path.is_empty() {
+                    span {
+                        style: "font-size: 12px; color: var(--text-secondary); font-family: monospace;",
+                        "{file_path}"
+                        if line_number > 0 {
+                            ":{line_number}"
+                        }
                     }
                 }
+                CopyButton { value: code.clone(), small: true }
             }
             pre { class: "code-block", "{code}" }
         }

compliance-dashboard/src/components/copy_button.rs

@@ -0,0 +1,49 @@
+use dioxus::prelude::*;
+use dioxus_free_icons::icons::bs_icons::*;
+use dioxus_free_icons::Icon;
+
+/// A small copy-to-clipboard button that shows a checkmark after copying.
+///
+/// Usage: `CopyButton { value: "text to copy" }`
+#[component]
+pub fn CopyButton(value: String, #[props(default = false)] small: bool) -> Element {
+    let mut copied = use_signal(|| false);
+
+    let size = if small { 12 } else { 14 };
+    let class = if small {
+        "copy-btn copy-btn-sm"
+    } else {
+        "copy-btn"
+    };
+
+    rsx! {
+        button {
+            class: class,
+            title: if copied() { "Copied!" } else { "Copy to clipboard" },
+            onclick: move |_| {
+                let val = value.clone();
+                // Escape for JS single-quoted string
+                let escaped = val
+                    .replace('\\', "\\\\")
+                    .replace('\'', "\\'")
+                    .replace('\n', "\\n")
+                    .replace('\r', "\\r");
+                let js = format!("navigator.clipboard.writeText('{escaped}')");
+                document::eval(&js);
+                copied.set(true);
+                spawn(async move {
+                    #[cfg(feature = "web")]
+                    gloo_timers::future::TimeoutFuture::new(2000).await;
+                    #[cfg(not(feature = "web"))]
+                    tokio::time::sleep(std::time::Duration::from_secs(2)).await;
+                    copied.set(false);
+                });
+            },
+            if copied() {
+                Icon { icon: BsCheckLg, width: size, height: size }
+            } else {
+                Icon { icon: BsClipboard, width: size, height: size }
+            }
+        }
+    }
+}

compliance-dashboard/src/components/help_chat.rs

@@ -0,0 +1,198 @@
+use dioxus::prelude::*;
+use dioxus_free_icons::icons::bs_icons::*;
+use dioxus_free_icons::Icon;
+
+use crate::infrastructure::help_chat::{send_help_chat_message, HelpChatHistoryMessage};
+
+// ── Message model ────────────────────────────────────────────────────────────
+
+#[derive(Clone, Debug)]
+struct ChatMsg {
+    role: String,
+    content: String,
+}
+
+// ── Component ────────────────────────────────────────────────────────────────
+
+#[component]
+pub fn HelpChat() -> Element {
+    let mut is_open = use_signal(|| false);
+    let mut messages = use_signal(Vec::<ChatMsg>::new);
+    let mut input_text = use_signal(String::new);
+    let mut is_loading = use_signal(|| false);
+
+    // Send message handler
+    let on_send = move |_| {
+        let text = input_text().trim().to_string();
+        if text.is_empty() || is_loading() {
+            return;
+        }
+
+        // Push user message
+        messages.write().push(ChatMsg {
+            role: "user".into(),
+            content: text.clone(),
+        });
+        input_text.set(String::new());
+        is_loading.set(true);
+
+        // Build history for API call (exclude last user message, it goes as `message`)
+        let history: Vec<HelpChatHistoryMessage> = messages()
+            .iter()
+            .rev()
+            .skip(1) // skip the user message we just added
+            .rev()
+            .map(|m| HelpChatHistoryMessage {
+                role: m.role.clone(),
+                content: m.content.clone(),
+            })
+            .collect();
+
+        spawn(async move {
+            match send_help_chat_message(text, history).await {
+                Ok(resp) => {
+                    messages.write().push(ChatMsg {
+                        role: "assistant".into(),
+                        content: resp.data.message,
+                    });
+                }
+                Err(e) => {
+                    messages.write().push(ChatMsg {
+                        role: "assistant".into(),
+                        content: format!("Error: {e}"),
+                    });
+                }
+            }
+            is_loading.set(false);
+        });
+    };
+
+    // Key handler for Enter to send
+    let on_keydown = move |e: KeyboardEvent| {
+        if e.key() == Key::Enter && !e.modifiers().shift() {
+            e.prevent_default();
+            let text = input_text().trim().to_string();
+            if text.is_empty() || is_loading() {
+                return;
+            }
+            messages.write().push(ChatMsg {
+                role: "user".into(),
+                content: text.clone(),
+            });
+            input_text.set(String::new());
+            is_loading.set(true);
+
+            let history: Vec<HelpChatHistoryMessage> = messages()
+                .iter()
+                .rev()
+                .skip(1)
+                .rev()
+                .map(|m| HelpChatHistoryMessage {
+                    role: m.role.clone(),
+                    content: m.content.clone(),
+                })
+                .collect();
+
+            spawn(async move {
+                match send_help_chat_message(text, history).await {
+                    Ok(resp) => {
+                        messages.write().push(ChatMsg {
+                            role: "assistant".into(),
+                            content: resp.data.message,
+                        });
+                    }
+                    Err(e) => {
+                        messages.write().push(ChatMsg {
+                            role: "assistant".into(),
+                            content: format!("Error: {e}"),
+                        });
+                    }
+                }
+                is_loading.set(false);
+            });
+        }
+    };
+
+    rsx! {
+        // Floating toggle button
+        if !is_open() {
+            button {
+                class: "help-chat-toggle",
+                onclick: move |_| is_open.set(true),
+                title: "Help",
+                Icon { icon: BsQuestionCircle, width: 22, height: 22 }
+            }
+        }
+
+        // Chat panel
+        if is_open() {
+            div { class: "help-chat-panel",
+                // Header
+                div { class: "help-chat-header",
+                    span { class: "help-chat-title",
+                        Icon { icon: BsRobot, width: 16, height: 16 }
+                        "Help Assistant"
+                    }
+                    button {
+                        class: "help-chat-close",
+                        onclick: move |_| is_open.set(false),
+                        Icon { icon: BsX, width: 18, height: 18 }
+                    }
+                }
+
+                // Messages area
+                div { class: "help-chat-messages",
+                    if messages().is_empty() {
+                        div { class: "help-chat-empty",
+                            p { "Ask me anything about the Compliance Scanner." }
+                            p { class: "help-chat-hint",
+                                "e.g. \"How do I add a repository?\" or \"What is SBOM?\""
+                            }
+                        }
+                    }
+                    for (i, msg) in messages().iter().enumerate() {
+                        div {
+                            key: "{i}",
+                            class: if msg.role == "user" { "help-msg help-msg-user" } else { "help-msg help-msg-assistant" },
+                            div { class: "help-msg-content",
+                                dangerous_inner_html: if msg.role == "assistant" {
+                                    // Basic markdown rendering: bold, code, newlines
+                                    msg.content
+                                        .replace("**", "<strong>")
+                                        .replace("\n\n", "<br><br>")
+                                        .replace("\n- ", "<br>- ")
+                                        .replace("`", "<code>")
+                                } else {
+                                    msg.content.clone()
+                                }
+                            }
+                        }
+                    }
+                    if is_loading() {
+                        div { class: "help-msg help-msg-assistant",
+                            div { class: "help-msg-loading", "Thinking..." }
+                        }
+                    }
+                }
+
+                // Input area
+                div { class: "help-chat-input",
+                    input {
+                        r#type: "text",
+                        placeholder: "Ask a question...",
+                        value: "{input_text}",
+                        disabled: is_loading(),
+                        oninput: move |e| input_text.set(e.value()),
+                        onkeydown: on_keydown,
+                    }
+                    button {
+                        class: "help-chat-send",
+                        disabled: is_loading() || input_text().trim().is_empty(),
+                        onclick: on_send,
+                        Icon { icon: BsSend, width: 14, height: 14 }
+                    }
+                }
+            }
+        }
+    }
+}

compliance-dashboard/src/components/mod.rs

@@ -2,11 +2,15 @@ pub mod app_shell;
 pub mod attack_chain;
 pub mod code_inspector;
 pub mod code_snippet;
+pub mod copy_button;
 pub mod file_tree;
+pub mod help_chat;
+pub mod notification_bell;
 pub mod page_header;
 pub mod pagination;
 pub mod pentest_wizard;
 pub mod severity_badge;
 pub mod sidebar;
 pub mod stat_card;
+pub mod theme_toggle;
 pub mod toast;

compliance-dashboard/src/components/notification_bell.rs

@@ -0,0 +1,155 @@
+use dioxus::prelude::*;
+use dioxus_free_icons::icons::bs_icons::*;
+use dioxus_free_icons::Icon;
+
+use crate::infrastructure::notifications::{
+    dismiss_notification, fetch_notification_count, fetch_notifications,
+    mark_all_notifications_read,
+};
+
+#[component]
+pub fn NotificationBell() -> Element {
+    let mut is_open = use_signal(|| false);
+    let mut count = use_signal(|| 0u64);
+    let mut notifications = use_signal(Vec::new);
+    let mut is_loading = use_signal(|| false);
+
+    // Poll notification count every 30 seconds
+    use_resource(move || async move {
+        loop {
+            if let Ok(c) = fetch_notification_count().await {
+                count.set(c);
+            }
+            #[cfg(feature = "web")]
+            {
+                gloo_timers::future::TimeoutFuture::new(30_000).await;
+            }
+            #[cfg(not(feature = "web"))]
+            {
+                tokio::time::sleep(std::time::Duration::from_secs(30)).await;
+            }
+        }
+    });
+
+    // Load notifications when panel opens
+    let load_notifications = move |_| {
+        is_open.set(!is_open());
+        if !is_open() {
+            return;
+        }
+        is_loading.set(true);
+        spawn(async move {
+            if let Ok(resp) = fetch_notifications().await {
+                notifications.set(resp.data);
+            }
+            // Mark all as read when panel opens
+            let _ = mark_all_notifications_read().await;
+            count.set(0);
+            is_loading.set(false);
+        });
+    };
+
+    let on_dismiss = move |id: String| {
+        spawn(async move {
+            let _ = dismiss_notification(id.clone()).await;
+            notifications.write().retain(|n| {
+                n.id.as_ref()
+                    .and_then(|v| v.get("$oid"))
+                    .and_then(|v| v.as_str())
+                    != Some(&id)
+            });
+        });
+    };
+
+    rsx! {
+        div { class: "notification-bell-wrapper",
+            // Bell button
+            button {
+                class: "notification-bell-btn",
+                onclick: load_notifications,
+                title: "CVE Alerts",
+                Icon { icon: BsBell, width: 18, height: 18 }
+                if count() > 0 {
+                    span { class: "notification-badge", "{count()}" }
+                }
+            }
+
+            // Dropdown panel
+            if is_open() {
+                div { class: "notification-panel",
+                    div { class: "notification-panel-header",
+                        span { "CVE Alerts" }
+                        button {
+                            class: "notification-close-btn",
+                            onclick: move |_| is_open.set(false),
+                            Icon { icon: BsX, width: 16, height: 16 }
+                        }
+                    }
+                    div { class: "notification-panel-body",
+                        if is_loading() {
+                            div { class: "notification-loading", "Loading..." }
+                        } else if notifications().is_empty() {
+                            div { class: "notification-empty",
+                                Icon { icon: BsShieldCheck, width: 32, height: 32 }
+                                p { "No CVE alerts" }
+                            }
+                        } else {
+                            for notif in notifications().iter() {
+                                {
+                                    let id = notif.id.as_ref()
+                                        .and_then(|v| v.get("$oid"))
+                                        .and_then(|v| v.as_str())
+                                        .unwrap_or("")
+                                        .to_string();
+                                    let sev_class = match notif.severity.as_str() {
+                                        "critical" => "sev-critical",
+                                        "high" => "sev-high",
+                                        "medium" => "sev-medium",
+                                        _ => "sev-low",
+                                    };
+                                    let dismiss_id = id.clone();
+                                    rsx! {
+                                        div { class: "notification-item",
+                                            div { class: "notification-item-header",
+                                                span { class: "notification-sev {sev_class}",
+                                                    "{notif.severity.to_uppercase()}"
+                                                }
+                                                span { class: "notification-cve-id",
+                                                    if let Some(ref url) = notif.url {
+                                                        a { href: "{url}", target: "_blank", "{notif.cve_id}" }
+                                                    } else {
+                                                        "{notif.cve_id}"
+                                                    }
+                                                }
+                                                if let Some(score) = notif.cvss_score {
+                                                    span { class: "notification-cvss", "CVSS {score:.1}" }
+                                                }
+                                                button {
+                                                    class: "notification-dismiss-btn",
+                                                    title: "Dismiss",
+                                                    onclick: move |_| on_dismiss(dismiss_id.clone()),
+                                                    Icon { icon: BsXCircle, width: 14, height: 14 }
+                                                }
+                                            }
+                                            div { class: "notification-item-pkg",
+                                                "{notif.package_name} {notif.package_version}"
+                                            }
+                                            div { class: "notification-item-repo",
+                                                "{notif.repo_name}"
+                                            }
+                                            if let Some(ref summary) = notif.summary {
+                                                div { class: "notification-item-summary",
+                                                    "{summary}"
+                                                }
+                                            }
+                                        }
+                                    }
+                                }
+                            }
+                        }
+                    }
+                }
+            }
+        }
+    }
+}

compliance-dashboard/src/components/sidebar.rs

@@ -4,6 +4,7 @@ use dioxus_free_icons::icons::bs_icons::*;
 use dioxus_free_icons::Icon;
 
 use crate::app::Route;
+use crate::components::theme_toggle::ThemeToggle;
 
 struct NavItem {
     label: &'static str,
@@ -52,11 +53,6 @@ pub fn Sidebar() -> Element {
             route: Route::PentestDashboardPage {},
             icon: rsx! { Icon { icon: BsLightningCharge, width: 18, height: 18 } },
         },
-        NavItem {
-            label: "Settings",
-            route: Route::SettingsPage {},
-            icon: rsx! { Icon { icon: BsGear, width: 18, height: 18 } },
-        },
     ];
 
     let docs_url = option_env!("DOCS_URL").unwrap_or("/docs");
@@ -111,6 +107,7 @@ pub fn Sidebar() -> Element {
             }
             // Spacer pushes footer to the bottom
             div { class: "sidebar-spacer" }
+            ThemeToggle { collapsed: collapsed() }
             button {
                 class: "sidebar-toggle",
                 onclick: move |_| collapsed.set(!collapsed()),

compliance-dashboard/src/components/theme_toggle.rs

@@ -0,0 +1,104 @@
+use dioxus::prelude::*;
+use dioxus_free_icons::icons::bs_icons::{BsMoonStars, BsSun};
+use dioxus_free_icons::Icon;
+
+#[cfg(feature = "web")]
+const STORAGE_KEY: &str = "compliance-scanner.theme";
+
+/// Sidebar-footer theme toggle. Reads the initial state on mount from
+/// localStorage (explicit user choice) or `prefers-color-scheme` (OS default),
+/// then writes back to both the `<html data-theme="...">` attribute and
+/// localStorage on every click.
+#[component]
+pub fn ThemeToggle(collapsed: bool) -> Element {
+    // `None` until the on-mount effect resolves the real value, so SSR doesn't
+    // render the wrong icon for the user's actual theme.
+    let mut is_dark = use_signal(|| None::<bool>);
+
+    use_effect(move || {
+        let (dark, from_storage) = initial_theme();
+        is_dark.set(Some(dark));
+        // If the user already made an explicit choice (in localStorage), assert it
+        // on the DOM so an OS-vs-stored mismatch can't briefly show the wrong theme.
+        if from_storage {
+            apply_theme(dark);
+        }
+    });
+
+    let label = if collapsed {
+        ""
+    } else if is_dark().unwrap_or(true) {
+        "Light mode"
+    } else {
+        "Dark mode"
+    };
+
+    let title = if is_dark().unwrap_or(true) {
+        "Switch to light mode"
+    } else {
+        "Switch to dark mode"
+    };
+
+    rsx! {
+        button {
+            class: "theme-toggle",
+            r#type: "button",
+            title: "{title}",
+            "aria-label": "{title}",
+            onclick: move |_| {
+                let next_dark = !is_dark().unwrap_or(true);
+                is_dark.set(Some(next_dark));
+                apply_theme(next_dark);
+            },
+            if is_dark().unwrap_or(true) {
+                Icon { icon: BsSun, width: 16, height: 16 }
+            } else {
+                Icon { icon: BsMoonStars, width: 16, height: 16 }
+            }
+            if !collapsed {
+                span { class: "theme-toggle-label", "{label}" }
+            }
+        }
+    }
+}
+
+/// Returns `(is_dark, from_storage)`. `from_storage` is true when an explicit
+/// user choice is in localStorage; false when we fell back to OS preference
+/// (or to the dark default).
+#[cfg(feature = "web")]
+fn initial_theme() -> (bool, bool) {
+    if let Some(window) = web_sys::window() {
+        if let Ok(Some(storage)) = window.local_storage() {
+            if let Ok(Some(value)) = storage.get_item(STORAGE_KEY) {
+                return (value == "dark", true);
+            }
+        }
+        if let Ok(Some(mql)) = window.match_media("(prefers-color-scheme: dark)") {
+            return (mql.matches(), false);
+        }
+    }
+    (true, false)
+}
+
+#[cfg(not(feature = "web"))]
+fn initial_theme() -> (bool, bool) {
+    (true, false)
+}
+
+#[cfg(feature = "web")]
+fn apply_theme(dark: bool) {
+    let theme = if dark { "dark" } else { "light" };
+    if let Some(window) = web_sys::window() {
+        if let Some(document) = window.document() {
+            if let Some(root) = document.document_element() {
+                let _ = root.set_attribute("data-theme", theme);
+            }
+        }
+        if let Ok(Some(storage)) = window.local_storage() {
+            let _ = storage.set_item(STORAGE_KEY, theme);
+        }
+    }
+}
+
+#[cfg(not(feature = "web"))]
+fn apply_theme(_dark: bool) {}

compliance-dashboard/src/infrastructure/agent_client.rs

@@ -0,0 +1,210 @@
+//! Authenticated HTTP client for talking to the compliance-agent.
+//!
+//! Every dashboard server function that hits `comp-dev.meghsakha.com/api/v1/*`
+//! must go through here so the Keycloak access token from the user's
+//! session is attached as `Authorization: Bearer <token>`. Without it
+//! the agent's M7.1 `require_jwt_auth` middleware rejects with 401
+//! "Missing authorization header".
+//!
+//! When Keycloak is not configured (dev convenience), the helper
+//! returns an unauthenticated builder — matching the agent's
+//! pass-through behavior in the same state.
+//!
+//! **Token refresh**: KC access tokens are short-lived (5 min default
+//! in the certifai realm). Before attaching, we decode the JWT's `exp`
+//! claim and proactively refresh via the stored refresh_token if the
+//! access token is expired or about to expire. The session is updated
+//! with the new pair. If refresh fails, we send the (stale) token
+//! anyway — the agent's 401 will surface to the UI, which can prompt
+//! re-login.
+
+use base64::{engine::general_purpose::URL_SAFE_NO_PAD, Engine};
+use dioxus::prelude::ServerFnError;
+use dioxus_fullstack::FullstackContext;
+use reqwest::Method;
+
+use super::auth::LOGGED_IN_USER_SESS_KEY;
+use super::server_state::ServerState;
+use super::user_state::UserStateInner;
+
+/// Seconds before the JWT's `exp` time at which we consider it stale
+/// enough to refresh. Covers clock skew + the round-trip to the agent
+/// so the token doesn't expire mid-flight.
+const REFRESH_SKEW_SECS: i64 = 30;
+
+/// Build a `RequestBuilder` for `<agent_api_url><path>` with the
+/// session's access token attached. `path` should include a leading
+/// `/`, e.g. `"/api/v1/repositories"`.
+pub async fn agent_request(
+    method: Method,
+    path: &str,
+) -> Result<reqwest::RequestBuilder, ServerFnError> {
+    let state: ServerState = FullstackContext::extract().await?;
+    let url = format!("{}{}", state.agent_api_url, path);
+    let mut req = reqwest::Client::new().request(method, &url);
+    req = attach_token(req, &state).await?;
+    Ok(req)
+}
+
+/// Same as [`agent_request`] but for `GET`. Convenience for the common case.
+pub async fn agent_get(path: &str) -> Result<reqwest::RequestBuilder, ServerFnError> {
+    agent_request(Method::GET, path).await
+}
+
+/// Attach the session's bearer token if Keycloak is configured AND the
+/// session has a logged-in user. Refresh the token proactively if it's
+/// expired or about to expire. Persists refreshed tokens back into the
+/// session.
+async fn attach_token(
+    req: reqwest::RequestBuilder,
+    state: &ServerState,
+) -> Result<reqwest::RequestBuilder, ServerFnError> {
+    if state.keycloak.is_none() {
+        return Ok(req);
+    }
+    let session: tower_sessions::Session = FullstackContext::extract().await?;
+    let user: Option<UserStateInner> = session
+        .get(LOGGED_IN_USER_SESS_KEY)
+        .await
+        .map_err(|e| ServerFnError::new(format!("session read failed: {e}")))?;
+    let Some(mut user) = user else {
+        return Ok(req);
+    };
+
+    if token_needs_refresh(&user.access_token) {
+        tracing::debug!("Access token expired or near-expiring; refreshing");
+        match refresh_tokens(state, &user.refresh_token).await {
+            Ok((new_access, new_refresh)) => {
+                user.access_token = new_access;
+                if let Some(rt) = new_refresh {
+                    user.refresh_token = rt;
+                }
+                if let Err(e) = session.insert(LOGGED_IN_USER_SESS_KEY, &user).await {
+                    tracing::warn!("Failed to persist refreshed tokens: {e}");
+                }
+            }
+            Err(e) => {
+                tracing::warn!("Token refresh failed: {e}; sending current token anyway");
+                // Fall through — the agent will 401 and the UI will
+                // prompt re-login. Better than failing the request at
+                // the dashboard layer with no helpful UX cue.
+            }
+        }
+    }
+
+    Ok(req.bearer_auth(user.access_token))
+}
+
+/// Decode the JWT's payload (no signature verification — the agent
+/// does that) and check the `exp` claim. Treats malformed tokens as
+/// expired so the refresh path runs.
+fn token_needs_refresh(jwt: &str) -> bool {
+    let Some(payload_b64) = jwt.split('.').nth(1) else {
+        return true;
+    };
+    let Ok(bytes) = URL_SAFE_NO_PAD.decode(payload_b64) else {
+        return true;
+    };
+    #[derive(serde::Deserialize)]
+    struct ExpClaim {
+        exp: i64,
+    }
+    let Ok(claims) = serde_json::from_slice::<ExpClaim>(&bytes) else {
+        return true;
+    };
+    let now = chrono::Utc::now().timestamp();
+    claims.exp - REFRESH_SKEW_SECS <= now
+}
+
+/// Exchange a refresh_token for a new access_token. Returns the new
+/// access_token and (optionally) the new refresh_token KC issued.
+/// KC may rotate refresh_tokens on use; we honor whatever it sends.
+async fn refresh_tokens(
+    state: &ServerState,
+    refresh_token: &str,
+) -> Result<(String, Option<String>), String> {
+    let kc = state
+        .keycloak
+        .ok_or_else(|| "Keycloak not configured".to_string())?;
+    if refresh_token.is_empty() {
+        return Err("no refresh_token in session".to_string());
+    }
+
+    #[derive(serde::Deserialize)]
+    struct TokenResp {
+        access_token: String,
+        refresh_token: Option<String>,
+    }
+
+    let resp = reqwest::Client::new()
+        .post(kc.token_endpoint())
+        .form(&[
+            ("grant_type", "refresh_token"),
+            ("client_id", kc.client_id.as_str()),
+            ("refresh_token", refresh_token),
+        ])
+        .send()
+        .await
+        .map_err(|e| format!("refresh request failed: {e}"))?;
+
+    if !resp.status().is_success() {
+        let status = resp.status();
+        let body = resp.text().await.unwrap_or_default();
+        return Err(format!("refresh rejected ({status}): {body}"));
+    }
+
+    let r: TokenResp = resp
+        .json()
+        .await
+        .map_err(|e| format!("refresh response parse failed: {e}"))?;
+    Ok((r.access_token, r.refresh_token))
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use base64::Engine;
+
+    /// Build a JWT-shaped string (header.payload.sig) with the given
+    /// payload. Signature is bogus — we never verify it locally.
+    fn make_jwt(payload: &serde_json::Value) -> String {
+        let payload_b64 = URL_SAFE_NO_PAD.encode(serde_json::to_vec(payload).unwrap());
+        format!("hdr.{payload_b64}.sig")
+    }
+
+    #[test]
+    fn token_needs_refresh_true_when_expired() {
+        let exp = chrono::Utc::now().timestamp() - 60;
+        let jwt = make_jwt(&serde_json::json!({ "exp": exp }));
+        assert!(token_needs_refresh(&jwt));
+    }
+
+    #[test]
+    fn token_needs_refresh_true_within_skew_window() {
+        // 10 seconds left; less than the 30s skew → must refresh.
+        let exp = chrono::Utc::now().timestamp() + 10;
+        let jwt = make_jwt(&serde_json::json!({ "exp": exp }));
+        assert!(token_needs_refresh(&jwt));
+    }
+
+    #[test]
+    fn token_needs_refresh_false_with_plenty_of_life() {
+        let exp = chrono::Utc::now().timestamp() + 600;
+        let jwt = make_jwt(&serde_json::json!({ "exp": exp }));
+        assert!(!token_needs_refresh(&jwt));
+    }
+
+    #[test]
+    fn token_needs_refresh_true_on_malformed_jwt() {
+        assert!(token_needs_refresh(""));
+        assert!(token_needs_refresh("not.a.jwt"));
+        assert!(token_needs_refresh("only-one-segment"));
+        assert!(token_needs_refresh("hdr.not-base64!.sig"));
+    }
+
+    #[test]
+    fn token_needs_refresh_true_when_exp_missing() {
+        let jwt = make_jwt(&serde_json::json!({ "sub": "abc" }));
+        assert!(token_needs_refresh(&jwt));
+    }
+}

compliance-dashboard/src/infrastructure/chat.rs

@@ -61,23 +61,21 @@ pub async fn send_chat_message(
     message: String,
     history: Vec<ChatHistoryMessage>,
 ) -> Result<ChatApiResponse, ServerFnError> {
-    let state: super::server_state::ServerState =
-        dioxus_fullstack::FullstackContext::extract().await?;
-
-    let url = format!("{}/api/v1/chat/{repo_id}", state.agent_api_url);
-    let client = reqwest::Client::builder()
-        .timeout(std::time::Duration::from_secs(120))
-        .build()
-        .map_err(|e| ServerFnError::new(e.to_string()))?;
-    let resp = client
-        .post(&url)
-        .json(&serde_json::json!({
-            "message": message,
-            "history": history,
-        }))
-        .send()
-        .await
-        .map_err(|e| ServerFnError::new(format!("Request failed: {e}")))?;
+    // Chat uses a longer timeout because the LLM round-trip can be slow;
+    // agent_request doesn't expose a per-call timeout so we layer one on.
+    let resp = super::agent_client::agent_request(
+        reqwest::Method::POST,
+        &format!("/api/v1/chat/{repo_id}"),
+    )
+    .await?
+    .timeout(std::time::Duration::from_secs(120))
+    .json(&serde_json::json!({
+        "message": message,
+        "history": history,
+    }))
+    .send()
+    .await
+    .map_err(|e| ServerFnError::new(format!("Request failed: {e}")))?;
 
     let text = resp
         .text()
@@ -91,19 +89,14 @@ pub async fn send_chat_message(
 
 #[server]
 pub async fn trigger_embedding_build(repo_id: String) -> Result<(), ServerFnError> {
-    let state: super::server_state::ServerState =
-        dioxus_fullstack::FullstackContext::extract().await?;
-
-    let url = format!(
-        "{}/api/v1/chat/{repo_id}/build-embeddings",
-        state.agent_api_url
-    );
-    let client = reqwest::Client::new();
-    client
-        .post(&url)
-        .send()
-        .await
-        .map_err(|e| ServerFnError::new(e.to_string()))?;
+    super::agent_client::agent_request(
+        reqwest::Method::POST,
+        &format!("/api/v1/chat/{repo_id}/build-embeddings"),
+    )
+    .await?
+    .send()
+    .await
+    .map_err(|e| ServerFnError::new(e.to_string()))?;
     Ok(())
 }
 
@@ -111,11 +104,9 @@ pub async fn trigger_embedding_build(repo_id: String) -> Result<(), ServerFnErro
 pub async fn fetch_embedding_status(
     repo_id: String,
 ) -> Result<EmbeddingStatusResponse, ServerFnError> {
-    let state: super::server_state::ServerState =
-        dioxus_fullstack::FullstackContext::extract().await?;
-
-    let url = format!("{}/api/v1/chat/{repo_id}/status", state.agent_api_url);
-    let resp = reqwest::get(&url)
+    let resp = super::agent_client::agent_get(&format!("/api/v1/chat/{repo_id}/status"))
+        .await?
+        .send()
         .await
         .map_err(|e| ServerFnError::new(e.to_string()))?;
     let body: EmbeddingStatusResponse = resp

compliance-dashboard/src/infrastructure/dast.rs

@@ -26,10 +26,9 @@ pub struct DastFindingDetailResponse {
 
 #[server]
 pub async fn fetch_dast_targets() -> Result<DastTargetsResponse, ServerFnError> {
-    let state: super::server_state::ServerState =
-        dioxus_fullstack::FullstackContext::extract().await?;
-    let url = format!("{}/api/v1/dast/targets", state.agent_api_url);
-    let resp = reqwest::get(&url)
+    let resp = super::agent_client::agent_get("/api/v1/dast/targets")
+        .await?
+        .send()
         .await
         .map_err(|e| ServerFnError::new(e.to_string()))?;
     let body: DastTargetsResponse = resp
@@ -41,10 +40,9 @@ pub async fn fetch_dast_targets() -> Result<DastTargetsResponse, ServerFnError>
 
 #[server]
 pub async fn fetch_dast_scan_runs() -> Result<DastScanRunsResponse, ServerFnError> {
-    let state: super::server_state::ServerState =
-        dioxus_fullstack::FullstackContext::extract().await?;
-    let url = format!("{}/api/v1/dast/scan-runs", state.agent_api_url);
-    let resp = reqwest::get(&url)
+    let resp = super::agent_client::agent_get("/api/v1/dast/scan-runs")
+        .await?
+        .send()
         .await
         .map_err(|e| ServerFnError::new(e.to_string()))?;
     let body: DastScanRunsResponse = resp
@@ -56,10 +54,9 @@ pub async fn fetch_dast_scan_runs() -> Result<DastScanRunsResponse, ServerFnErro
 
 #[server]
 pub async fn fetch_dast_findings() -> Result<DastFindingsResponse, ServerFnError> {
-    let state: super::server_state::ServerState =
-        dioxus_fullstack::FullstackContext::extract().await?;
-    let url = format!("{}/api/v1/dast/findings", state.agent_api_url);
-    let resp = reqwest::get(&url)
+    let resp = super::agent_client::agent_get("/api/v1/dast/findings")
+        .await?
+        .send()
         .await
         .map_err(|e| ServerFnError::new(e.to_string()))?;
     let body: DastFindingsResponse = resp
@@ -73,10 +70,9 @@ pub async fn fetch_dast_findings() -> Result<DastFindingsResponse, ServerFnError
 pub async fn fetch_dast_finding_detail(
     id: String,
 ) -> Result<DastFindingDetailResponse, ServerFnError> {
-    let state: super::server_state::ServerState =
-        dioxus_fullstack::FullstackContext::extract().await?;
-    let url = format!("{}/api/v1/dast/findings/{id}", state.agent_api_url);
-    let resp = reqwest::get(&url)
+    let resp = super::agent_client::agent_get(&format!("/api/v1/dast/findings/{id}"))
+        .await?
+        .send()
         .await
         .map_err(|e| ServerFnError::new(e.to_string()))?;
     let body: DastFindingDetailResponse = resp
@@ -88,12 +84,8 @@ pub async fn fetch_dast_finding_detail(
 
 #[server]
 pub async fn add_dast_target(name: String, base_url: String) -> Result<(), ServerFnError> {
-    let state: super::server_state::ServerState =
-        dioxus_fullstack::FullstackContext::extract().await?;
-    let url = format!("{}/api/v1/dast/targets", state.agent_api_url);
-    let client = reqwest::Client::new();
-    client
-        .post(&url)
+    super::agent_client::agent_request(reqwest::Method::POST, "/api/v1/dast/targets")
+        .await?
         .json(&serde_json::json!({
             "name": name,
             "base_url": base_url,
@@ -106,17 +98,13 @@ pub async fn add_dast_target(name: String, base_url: String) -> Result<(), Serve
 
 #[server]
 pub async fn trigger_dast_scan(target_id: String) -> Result<(), ServerFnError> {
-    let state: super::server_state::ServerState =
-        dioxus_fullstack::FullstackContext::extract().await?;
-    let url = format!(
-        "{}/api/v1/dast/targets/{target_id}/scan",
-        state.agent_api_url
-    );
-    let client = reqwest::Client::new();
-    client
-        .post(&url)
-        .send()
-        .await
-        .map_err(|e| ServerFnError::new(e.to_string()))?;
+    super::agent_client::agent_request(
+        reqwest::Method::POST,
+        &format!("/api/v1/dast/targets/{target_id}/scan"),
+    )
+    .await?
+    .send()
+    .await
+    .map_err(|e| ServerFnError::new(e.to_string()))?;
     Ok(())
 }

compliance-dashboard/src/infrastructure/findings.rs

@@ -24,39 +24,35 @@ pub struct FindingsQuery {
 
 #[server]
 pub async fn fetch_findings(query: FindingsQuery) -> Result<FindingsListResponse, ServerFnError> {
-    let state: super::server_state::ServerState =
-        dioxus_fullstack::FullstackContext::extract().await?;
-
-    let mut url = format!(
-        "{}/api/v1/findings?page={}&limit=20",
-        state.agent_api_url, query.page
-    );
+    let mut path = format!("/api/v1/findings?page={}&limit=20", query.page);
     if !query.severity.is_empty() {
-        url.push_str(&format!("&severity={}", query.severity));
+        path.push_str(&format!("&severity={}", query.severity));
     }
     if !query.scan_type.is_empty() {
-        url.push_str(&format!("&scan_type={}", query.scan_type));
+        path.push_str(&format!("&scan_type={}", query.scan_type));
     }
     if !query.status.is_empty() {
-        url.push_str(&format!("&status={}", query.status));
+        path.push_str(&format!("&status={}", query.status));
     }
     if !query.repo_id.is_empty() {
-        url.push_str(&format!("&repo_id={}", query.repo_id));
+        path.push_str(&format!("&repo_id={}", query.repo_id));
     }
     if !query.q.is_empty() {
-        url.push_str(&format!(
+        path.push_str(&format!(
             "&q={}",
             url::form_urlencoded::byte_serialize(query.q.as_bytes()).collect::<String>()
         ));
     }
     if !query.sort_by.is_empty() {
-        url.push_str(&format!("&sort_by={}", query.sort_by));
+        path.push_str(&format!("&sort_by={}", query.sort_by));
     }
     if !query.sort_order.is_empty() {
-        url.push_str(&format!("&sort_order={}", query.sort_order));
+        path.push_str(&format!("&sort_order={}", query.sort_order));
     }
 
-    let resp = reqwest::get(&url)
+    let resp = super::agent_client::agent_get(&path)
+        .await?
+        .send()
         .await
         .map_err(|e| ServerFnError::new(e.to_string()))?;
     let body: FindingsListResponse = resp
@@ -68,11 +64,9 @@ pub async fn fetch_findings(query: FindingsQuery) -> Result<FindingsListResponse
 
 #[server]
 pub async fn fetch_finding_detail(id: String) -> Result<Finding, ServerFnError> {
-    let state: super::server_state::ServerState =
-        dioxus_fullstack::FullstackContext::extract().await?;
-    let url = format!("{}/api/v1/findings/{id}", state.agent_api_url);
-
-    let resp = reqwest::get(&url)
+    let resp = super::agent_client::agent_get(&format!("/api/v1/findings/{id}"))
+        .await?
+        .send()
         .await
         .map_err(|e| ServerFnError::new(e.to_string()))?;
     let body: serde_json::Value = resp
@@ -86,18 +80,15 @@ pub async fn fetch_finding_detail(id: String) -> Result<Finding, ServerFnError>
 
 #[server]
 pub async fn update_finding_status(id: String, status: String) -> Result<(), ServerFnError> {
-    let state: super::server_state::ServerState =
-        dioxus_fullstack::FullstackContext::extract().await?;
-    let url = format!("{}/api/v1/findings/{id}/status", state.agent_api_url);
-
-    let client = reqwest::Client::new();
-    client
-        .patch(&url)
-        .json(&serde_json::json!({ "status": status }))
-        .send()
-        .await
-        .map_err(|e| ServerFnError::new(e.to_string()))?;
-
+    super::agent_client::agent_request(
+        reqwest::Method::PATCH,
+        &format!("/api/v1/findings/{id}/status"),
+    )
+    .await?
+    .json(&serde_json::json!({ "status": status }))
+    .send()
+    .await
+    .map_err(|e| ServerFnError::new(e.to_string()))?;
     Ok(())
 }
 
@@ -106,34 +97,25 @@ pub async fn bulk_update_finding_status(
     ids: Vec<String>,
     status: String,
 ) -> Result<(), ServerFnError> {
-    let state: super::server_state::ServerState =
-        dioxus_fullstack::FullstackContext::extract().await?;
-    let url = format!("{}/api/v1/findings/bulk-status", state.agent_api_url);
-
-    let client = reqwest::Client::new();
-    client
-        .patch(&url)
+    super::agent_client::agent_request(reqwest::Method::PATCH, "/api/v1/findings/bulk-status")
+        .await?
         .json(&serde_json::json!({ "ids": ids, "status": status }))
         .send()
         .await
         .map_err(|e| ServerFnError::new(e.to_string()))?;
-
     Ok(())
 }
 
 #[server]
 pub async fn update_finding_feedback(id: String, feedback: String) -> Result<(), ServerFnError> {
-    let state: super::server_state::ServerState =
-        dioxus_fullstack::FullstackContext::extract().await?;
-    let url = format!("{}/api/v1/findings/{id}/feedback", state.agent_api_url);
-
-    let client = reqwest::Client::new();
-    client
-        .patch(&url)
-        .json(&serde_json::json!({ "feedback": feedback }))
-        .send()
-        .await
-        .map_err(|e| ServerFnError::new(e.to_string()))?;
-
+    super::agent_client::agent_request(
+        reqwest::Method::PATCH,
+        &format!("/api/v1/findings/{id}/feedback"),
+    )
+    .await?
+    .json(&serde_json::json!({ "feedback": feedback }))
+    .send()
+    .await
+    .map_err(|e| ServerFnError::new(e.to_string()))?;
     Ok(())
 }

compliance-dashboard/src/infrastructure/graph.rs

@@ -50,10 +50,9 @@ pub struct SearchResponse {
 
 #[server]
 pub async fn fetch_graph(repo_id: String) -> Result<GraphDataResponse, ServerFnError> {
-    let state: super::server_state::ServerState =
-        dioxus_fullstack::FullstackContext::extract().await?;
-    let url = format!("{}/api/v1/graph/{repo_id}", state.agent_api_url);
-    let resp = reqwest::get(&url)
+    let resp = super::agent_client::agent_get(&format!("/api/v1/graph/{repo_id}"))
+        .await?
+        .send()
         .await
         .map_err(|e| ServerFnError::new(e.to_string()))?;
     let body: GraphDataResponse = resp
@@ -68,15 +67,12 @@ pub async fn fetch_impact(
     repo_id: String,
     finding_id: String,
 ) -> Result<ImpactResponse, ServerFnError> {
-    let state: super::server_state::ServerState =
-        dioxus_fullstack::FullstackContext::extract().await?;
-    let url = format!(
-        "{}/api/v1/graph/{repo_id}/impact/{finding_id}",
-        state.agent_api_url
-    );
-    let resp = reqwest::get(&url)
-        .await
-        .map_err(|e| ServerFnError::new(e.to_string()))?;
+    let resp =
+        super::agent_client::agent_get(&format!("/api/v1/graph/{repo_id}/impact/{finding_id}"))
+            .await?
+            .send()
+            .await
+            .map_err(|e| ServerFnError::new(e.to_string()))?;
     let body: ImpactResponse = resp
         .json()
         .await
@@ -86,10 +82,9 @@ pub async fn fetch_impact(
 
 #[server]
 pub async fn fetch_communities(repo_id: String) -> Result<CommunitiesResponse, ServerFnError> {
-    let state: super::server_state::ServerState =
-        dioxus_fullstack::FullstackContext::extract().await?;
-    let url = format!("{}/api/v1/graph/{repo_id}/communities", state.agent_api_url);
-    let resp = reqwest::get(&url)
+    let resp = super::agent_client::agent_get(&format!("/api/v1/graph/{repo_id}/communities"))
+        .await?
+        .send()
         .await
         .map_err(|e| ServerFnError::new(e.to_string()))?;
     let body: CommunitiesResponse = resp
@@ -104,15 +99,13 @@ pub async fn fetch_file_content(
     repo_id: String,
     file_path: String,
 ) -> Result<FileContentResponse, ServerFnError> {
-    let state: super::server_state::ServerState =
-        dioxus_fullstack::FullstackContext::extract().await?;
-    let url = format!(
-        "{}/api/v1/graph/{repo_id}/file-content?path={file_path}",
-        state.agent_api_url
-    );
-    let resp = reqwest::get(&url)
-        .await
-        .map_err(|e| ServerFnError::new(e.to_string()))?;
+    let resp = super::agent_client::agent_get(&format!(
+        "/api/v1/graph/{repo_id}/file-content?path={file_path}"
+    ))
+    .await?
+    .send()
+    .await
+    .map_err(|e| ServerFnError::new(e.to_string()))?;
     let body: FileContentResponse = resp
         .json()
         .await
@@ -122,15 +115,13 @@ pub async fn fetch_file_content(
 
 #[server]
 pub async fn search_nodes(repo_id: String, query: String) -> Result<SearchResponse, ServerFnError> {
-    let state: super::server_state::ServerState =
-        dioxus_fullstack::FullstackContext::extract().await?;
-    let url = format!(
-        "{}/api/v1/graph/{repo_id}/search?q={query}&limit=50",
-        state.agent_api_url
-    );
-    let resp = reqwest::get(&url)
-        .await
-        .map_err(|e| ServerFnError::new(e.to_string()))?;
+    let resp = super::agent_client::agent_get(&format!(
+        "/api/v1/graph/{repo_id}/search?q={query}&limit=50"
+    ))
+    .await?
+    .send()
+    .await
+    .map_err(|e| ServerFnError::new(e.to_string()))?;
     let body: SearchResponse = resp
         .json()
         .await
@@ -140,14 +131,13 @@ pub async fn search_nodes(repo_id: String, query: String) -> Result<SearchRespon
 
 #[server]
 pub async fn trigger_graph_build(repo_id: String) -> Result<(), ServerFnError> {
-    let state: super::server_state::ServerState =
-        dioxus_fullstack::FullstackContext::extract().await?;
-    let url = format!("{}/api/v1/graph/{repo_id}/build", state.agent_api_url);
-    let client = reqwest::Client::new();
-    client
-        .post(&url)
-        .send()
-        .await
-        .map_err(|e| ServerFnError::new(e.to_string()))?;
+    super::agent_client::agent_request(
+        reqwest::Method::POST,
+        &format!("/api/v1/graph/{repo_id}/build"),
+    )
+    .await?
+    .send()
+    .await
+    .map_err(|e| ServerFnError::new(e.to_string()))?;
     Ok(())
 }

compliance-dashboard/src/infrastructure/help_chat.rs

@@ -0,0 +1,59 @@
+use dioxus::prelude::*;
+use serde::{Deserialize, Serialize};
+
+// ── Response types ──
+
+#[derive(Debug, Clone, Serialize, Deserialize, Default)]
+pub struct HelpChatApiResponse {
+    pub data: HelpChatResponseData,
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize, Default)]
+pub struct HelpChatResponseData {
+    pub message: String,
+}
+
+// ── History message type ──
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct HelpChatHistoryMessage {
+    pub role: String,
+    pub content: String,
+}
+
+// ── Server function ──
+
+#[server]
+pub async fn send_help_chat_message(
+    message: String,
+    history: Vec<HelpChatHistoryMessage>,
+) -> Result<HelpChatApiResponse, ServerFnError> {
+    let state: super::server_state::ServerState =
+        dioxus_fullstack::FullstackContext::extract().await?;
+
+    let url = format!("{}/api/v1/help/chat", state.agent_api_url);
+    let client = reqwest::Client::builder()
+        .timeout(std::time::Duration::from_secs(120))
+        .build()
+        .map_err(|e| ServerFnError::new(e.to_string()))?;
+
+    let resp = client
+        .post(&url)
+        .json(&serde_json::json!({
+            "message": message,
+            "history": history,
+        }))
+        .send()
+        .await
+        .map_err(|e| ServerFnError::new(format!("Help chat request failed: {e}")))?;
+
+    let text = resp
+        .text()
+        .await
+        .map_err(|e| ServerFnError::new(format!("Failed to read response: {e}")))?;
+
+    let body: HelpChatApiResponse = serde_json::from_str(&text)
+        .map_err(|e| ServerFnError::new(format!("Failed to parse response: {e}")))?;
+
+    Ok(body)
+}