The security layer your AI agent stack is missing.

A secure proxy between your agents and your databases. Send natural language; get safe SQL, per-agent row-level security, and full audit. Agents never see connection strings.

SELECT-only AST validator Per-agent RLS Append-only audit MCP-native

Sign up — get an API key in 30 seconds

Free tier: 3 databases · 1M queries/month · no credit card.

How it works

  1. You sign up; we hand you an admin API key + dashboard.
  2. You register your DB connection string. We encrypt it at rest with AES-128 (Fernet); your agents never see it.
  3. Your agent calls POST /v1/query in natural language. We translate to SQL via Claude, validate at the AST level, apply your row-level security policy, execute, and audit-log every call.

MCP integration

Drop into Claude Desktop / Cursor / any MCP-aware client:

pip install queryshield-mcp

# .mcp.json or claude_desktop_config.json
{
  "queryshield": {
    "command": "queryshield-mcp",
    "env": { "QUERYSHIELD_API_KEY": "qs_..." }
  }
}

Listed in the official MCP Registry as io.github.bch1212/queryshield.

Pricing

TierMonthlyDatabasesQueries/month
Starter$50031M
Pro$1,5001010M (audit export)
Enterprise$3,500unlimitedSSO + SIEM webhook

Docs: /docs · Health: /health · github.com/bch1212/queryshield