Missing row-level security
A new table enters an exposed schema without enabling or enforcing tenant policies.
BoundaryCI combines deterministic CI checks with optional managed AI review to catch dangerous Supabase and PostgreSQL authorization changes before they reach production.
npx boundaryci scan .Free deterministic scanner · Managed AI on paid plans · No database credentials
Scanning 12 SQL migrations...
BND004Every signed-in user can read rows without checking the active organization.
supabase/migrations/004_customers.sql:17using (true)Multi-tenant SaaS depends on small authorization policies being correct everywhere. One permissive migration can quietly turn a shared database into a cross-customer breach.
A new table enters an exposed schema without enabling or enforcing tenant policies.
An authenticated check proves a user signed in, but never proves which tenant they belong to.
A security-definer function introduces a path around otherwise-correct access controls.
Start locally in seconds. Add Cloud when your team needs durable proof across repositories.
LOCALBoundaryCI follows migrations in order and evaluates what security actually remains.
CIGitHub annotations point to the vulnerable policy before the pull request can merge.
CLOUDStore minimized findings, usage, and remediation evidence behind tenant-safe access.
BoundaryCI's high-confidence deterministic checks remain the reliable merge gate. Paid teams can authorize managed Fireworks AI to examine how tenant keys, memberships, RLS policies, and privileged functions work together—without managing another provider key.
Repeatable final-state rule with file evidence and remediation.
HIGH · BLOCKINGThe policy proves membership in an organization, but does not compare it with projects.organization_id.
Deterministic scans run inside your machine or GitHub runner. Cloud history receives a minimized, secret-redacted result. Paid teams can separately authorize managed AI review, which forwards locally redacted migration text without storing it.
Annual plans include two months free. Manage plan changes and schedule cancellation through Stripe.
Prove tenant isolation on your first SaaS repository.
Continuous protection for a shipping SaaS team.
Higher-volume assurance across a growing SaaS portfolio.
Paid subscriptions renew automatically until canceled. Cancellation normally takes effect at the end of the paid period. Charges are non-refundable except where required by law or agreed in writing. Stripe processes card information; BoundaryCI does not store complete card numbers. Review the subscription terms and privacy notice before purchase.
Create the repository in BoundaryCI Cloud, store its one-time token as the GitHub Actions secret BOUNDARYCI_CLOUD_TOKEN, and commit the generated workflow YAML. The token never goes in the file, and the repository dashboard keeps the safe YAML available under Setup guide.
Yes. An organization owner or administrator can select Add repository in the dashboard. Each repository receives its own bound token, setup guide, and scan history, while usage is counted against the organization's shared monthly allowance.
No. The current scanner analyzes migration files and does not require database credentials. Active testing against disposable environments is a future, separately configured capability.
Cloud history receives repository and commit context, summary counts, finding metadata, and short redacted evidence and remediation snippets. Complete migration files are excluded from history. If a paid organization separately enables managed AI, up to 80,000 characters of locally redacted migration text pass transiently through BoundaryCI to Fireworks and are not stored by BoundaryCI.
No. BoundaryCI's deterministic checks work without AI and remain the source of truth for merge decisions. Team, Growth, and Enterprise organizations can authorize BoundaryCI's managed Fireworks review without creating their own Fireworks account, then disable it by organization, repository, or workflow.
No. BoundaryCI is a focused continuous control for tenant-isolation regressions. It complements threat modeling, code review, testing, and independent security assessments.
Yes. Commit a reviewed baseline, then fail CI only when a new regression appears. Owned, expiring waivers keep temporary exceptions visible.
Run deterministic checks locally, then connect your repository for optional managed AI review and durable evidence.