Database★★★★4/5
PostgresMCPPro
Postgres MCP Pro is the most capable PostgreSQL MCP server available. It provides 9 tools covering query execution, schema exploration, health checks, and — most valuably — workload-based index recommendations using real query performance data from pg_stat_statements.
The index tuning alone is worth the setup. It analyzes your actual query patterns, recommends optimal indexes, and can simulate their impact with hypothetical index analysis in EXPLAIN plans. Configurable read-only or unrestricted access modes let you control the blast radius.
Available as a Docker image or Python package. SSE transport support means multiple clients can share one server instance.
Pros
- + 9 tools including workload-based index recommendations
- + Uses pg_stat_statements for real query performance data
- + Configurable read-only or unrestricted access modes
- + Hypothetical index simulation in EXPLAIN plans
- + SSE transport for multi-client sharing
Cons
- - Requires pg_stat_statements extension for full functionality
- - Index tuning features add complexity — overkill for simple queries
- - SSE transport still maturing
How We Use It
The nxsi system runs on a 50-table PostgreSQL database with 90 indexes, pgvector for embeddings, and analytics tables that grow daily. Postgres MCP Pro handles two things we used to do manually: schema exploration and index tuning.
For schema exploration — when Claude needs to write a query against the analytics tables or the task system, it can inspect the actual column types and constraints instead of relying on documentation that may be stale. The column naming inconsistencies across our tables (checked_at vs last_check_at, cost_usd vs estimated_cost) have bitten us enough times that having live schema access is worth the setup alone.
For index tuning — after the analytics system went live, query patterns shifted. The health check identified missing indexes on analytics_events that were causing sequential scans on the daily aggregation job. Would have caught it eventually in slow logs, but this was faster.
postgresdatabaseSQLperformanceindex-tuning