Database guardrails for AI agents that block destructive SQL before it reaches your database
Application security testing that proves what each account can access and exploits vulnerabilities with proof of fix
Stage
Startup
Not given
Founded
2025
Not given
Based in
Not given
IN
Pricing model
Not given
Paid
Pricing
Free plan with 500K queries/month, 1 database, and 7-day query history. Builder $79/month (billed $758/year) with 5M queries/month, 3 databases, 30-day history. Team $249/month (billed $2,390/year) with 50M queries/month and 90-day history. Enterprise from $2,500/month with unlimited queries, annual
Pilot: $199 for 30 days on one application (credited against annual if you convert). Founding annual: $699 per application per year, locked for two years, available until 28 February 2027. Private: $944 per application per year for self-hosted agent. Partner: $489 per client app per year for reselle
Key features
Block destructive SQL with suggested fixes before it reaches the database
Mask sensitive columns in query results
Per-agent allowlists for table and column access
Rule simulator to test changes against real traffic
TCP proxy for PostgreSQL and MySQL
MCP integration for AI coding agents
Ed25519-signed audit trail with exportable reports
42 documented standard rules across 6 SQL dialects
GitHub Actions and CI/CD integration
Command-line tool for offline SQL validation
Multi-role access control testing with real logins for different user types
Exploited vulnerability proof with request, response, screenshot and code line
Automated re-exploitation to prove fixes worked
Private agent execution inside customer network
Ten vulnerability classes including authorization and business logic abuse
Unlimited scans with bring-your-own AI key model
CI/CD integration with GitHub Actions for build gating
SSO, MFA, RBAC and audit log support
What makes it different
Deterministic AST-based rules: same statement always gets the same verdict, with no AI variance or false positives
Blocks destructive queries before they reach the database with automatic fix suggestions
Works across multiple SQL dialects and deployment modes without code changes to applications
Minimal latency overhead: 0.08 ms on PostgreSQL and 0.07 ms on MySQL at median
Open source proxy and engine under AGPL-3.0 with free production usage
Only reports vulnerabilities actually exploited, not suspected findings
Proves fixes work by re-running the same attack that previously succeeded
Tests authorization with multiple real user roles instead of single scanner login
Source code stays in customer's own network with zero inbound ports
Integrates code analysis with runtime testing for comprehensive vulnerability detection