Midplane
A thin access layer in front of existing Postgres that makes it safe for AI agents — read-only by default, writes opt-in per table, every query logged.
Where Midplane ranks
Current edition · updated Oct 20, 2026About Midplane
Companies are connecting AI coding agents like Cursor and Claude directly to their databases — where one wrong command can delete data or leak customer records. Midplane sits in between and controls what each agent is allowed to do: it blocks dangerous queries, hides sensitive data, and keeps a record of everything the agent touches. It’s open source, and runs either as a hosted service or on your own servers.
Products
A thin access layer / MCP server that sits in front of Postgres, enforcing query-level policies, column masking, and audit logging.
Security and compliance
As reported by Midplane. Ask for audit reports before you buy.
- AES-256-GCM encryption for database credentials via per-region AWS KMS keys with envelope encryption
- Machine tokens stored as HMAC-SHA256 digests with constant-time comparison
- Postgres row-level security for audit log tenant isolation
- Browser hardening headers: Content-Security-Policy (frame-ancestors 'none'), X-Frame-Options (DENY), X-Content-Type-Options (nosniff), Referrer-Policy (strict-origin-when-cross-origin), Permissions-Policy (camera/microphone/geolocation disabled), Strict-Transport-Security (2-year max-age, includeSubDomains, preload)
- Per-agent least-privilege scoping with fail-closed on no-grant credentials
- Per-region cryptographic key separation for credentials and tokens
- Strictly necessary cookies only; no advertising or third-party advertising trackers
Is this your company?
Claim the profile to keep facts current, answer buyers and submit evidence for the next ranking review.
Claim Midplane