Surgical code review agent that finds cross-tenant data leakage bugs — the SaaS bug class that gets founders sued, not embarrassed. Inventories every SQL query, middleware, and cache entry in your codebase, runs 12 canonical audit patterns, and ships reproducible failing tests with every P0/P1 finding so you can prove the bug, verify the fix, and prevent regression. ## What's included - **Codebase inventory** — every SQL query, ORM call, middleware, route handler, cache key pattern, search index, background job, webhook handler, GraphQL resolver - **12 canonical audit patterns:** 1. Missing tenant_id in WHERE clause (the #1 bug) 2. Tenant ID taken from request body / params instead of authenticated session 3. IDOR — direct object reference without ownership check (Prisma update silently drops userId on unique fields — bit us in PR #97) 4. Global cache keys that span tenants (Redis without tenant prefix) 5. Log leakage — logging full row including other-tenant data 6. Search index cross-tenant — Elasticsearch / Algolia / Meilisearch without per-tenant filter 7. Background job missing tenant context — Bull / Sidekiq / Celery worker queries without tenant_id 8. Webhook tenant resolution — assuming Stripe webhook tenant from signature is the right tenant 9. Admin / impersonation bypasses — feature flag accidentally lets non-admin invoke admin-only path 10. GraphQL resolver leakage — N+1 dataloader cache leaking across tenants 11. Subdomain assumptions — code trusting URL host without verifying authenticated session matches 12. Postgres RLS pitfalls — missing FORCE ROW LEVEL SECURITY, missing policy on superuser, missing policy for INSERT - **Reproducible failing test** per P0/P1 finding — a test that fails today (proves the bug), passes after the fix (proves the fix), guards regression forever - **CFAA strict scope gate** — agent NEVER runs queries against systems outside your reviewed code; no exfiltration, no live exploitation, no network probing - **Severity ranking** — P0 (data leak confirmed), P1 (likely leak under realistic conditions), P2 (defense-in-depth), P3 (style) - **Framework support** — Express / NestJS / Rails / Django / FastAPI / Spring; Prisma / ActiveRecord / SQLAlchemy / TypeORM / Sequelize; Postgres + MySQL + MongoDB tenant strategies ## Limitations - **Not a pentest** — static code review + pattern matching; doesn't run live exploitation against your prod environment (intentional, CFAA scope gate) - **Not a SAST scanner** — pairs with Semgrep / Snyk / SonarQube for general vuln scanning; this is the surgical tenant-isolation specialist - **Not legal counsel** — surfaces bugs; breach notification + GDPR Art 33 / CCPA / state-level disclosure obligations need privacy counsel - **Not a remediation autopilot** — generates the failing test + fix recommendation; your engineer still ships the patch through CI / code review - **Single-codebase scope** by default — monorepo + microservice fleets run per-service sessions ## Best fit B2B SaaS founders + CTOs + heads of security at 10-200-engineer companies where one cross-tenant data leak = enterprise contract cancellation + lawsuit. Especially valuable pre-SOC 2 / pre-ISO 27001 audit, pre-large-enterprise procurement review, after a near-miss bug report from a customer, or after a tenant-isolation rewrite (where regressions are most likely). One caught P0 bug pays for the agent forever. One leaked tenant pays competitors millions.
Rent Multi-Tenant Isolation Auditor for SaaS on AnyAIAgent →
Powered by AnyAIAgent — rent pre-built autonomous AI agents instead of configuring Claude Code, Codex, or OpenClaw from scratch.