{"id":"mgorunuch-postgres-mcp-server","name":"postgres-mcp-server","af_score":47.5,"security_score":39.8,"reliability_score":21.2,"what_it_does":"PostgreSQL MCP server that enables an MCP client (e.g., Claude Code) to query a PostgreSQL database using the MCP protocol. It can be run directly or configured via a generated JSON MCP config, and it blocks potentially destructive SQL statements by default unless explicitly marked unsafe.","best_when":"You want an MCP integration for a single Postgres connection and primarily read-only query workloads, with the agent constrained from destructive operations by default.","avoid_when":"You cannot provide a secure PostgreSQL connection/credentials (least-privilege), or you need strong guarantees around query approval, auditing, pagination, and operational resilience that are not documented here.","last_evaluated":"2026-04-04T21:21:56.818728+00:00","has_mcp":true,"has_api":false,"auth_methods":["PostgreSQL connection string provided via --connection-string or POSTGRES_CONNECTION_STRING env var"],"has_free_tier":false,"known_gotchas":["Destructive statements (DROP/TRUNCATE/DELETE/UPDATE/ALTER/CREATE/INSERT) are blocked by default; agents may need to avoid or explicitly set unsafe behavior (described as unsafe: true) if they truly require writes.","The safety model depends on the server/pg_query tool behavior and the agent adhering to the documented unsafe flag; otherwise write attempts may fail.","Large result sets may be returned without explicit pagination controls (not described)."],"error_quality":0.0}