DB Query Guard MCP

This is a paid remote MCP that sits between your agent and your database to enforce query policy before anything executes. It exposes tools like simulate_query, request_query_approval, and export_query_audit so agents can ask questions in natural language, generate SQL, and get a verdict (pass, review, or deny) based on rules like SELECT-only, row limits, and PII field detection. If a query touches sensitive columns or violates policy, it enters an approval queue where a human reviewer grants an exception with a reason and expiry. Every action produces an audit trail with query hashes and redacted previews. You'd reach for this when you want agents querying Postgres, Supabase, or MongoDB in production without handing them raw credentials or risking unreviewed access to customer data.