How to Use PostgreSQL MCP with Claude Code

What is the PostgreSQL MCP?

The PostgreSQL MCP creates a secure tunnel between Claude Code and any Postgres-compatible database. Unlike ORMs or GUI tools, it lets Claude issue raw SQL, inspect execution plans, and reason about your schema at the database level — ideal for performance tuning, schema design, and data debugging.

Installation

mcpizy install postgres

Provide your connection string (e.g., postgresql://user:password@host:5432/dbname) during setup. SSL connections are supported and recommended for remote databases.

Key Capabilities

• Execute arbitrary SQL — SELECT, INSERT, UPDATE, DELETE, and DDL statements.
• Inspect schemas — list tables, columns, indexes, constraints, and foreign keys.
• Analyze query plans — run EXPLAIN ANALYZE and let Claude interpret the output.
• Generate migrations — describe a schema change in plain English; Claude writes the SQL.
• Monitor slow queries — query pg_stat_statements and identify bottlenecks.

Example Usage

Identify the slowest queries in your production database:

SELECT query, calls, total_exec_time / calls AS avg_ms
FROM pg_stat_statements
ORDER BY avg_ms DESC
LIMIT 10;

Claude runs this via the MCP, then proposes missing indexes or query rewrites based on the results.

Tips and Best Practices

• Create a read-only role for exploratory queries and a write role only for approved migrations.
• Use BEGIN; ... ROLLBACK; blocks when testing destructive statements before committing.
• Pair with the Supabase MCP if your Postgres instance is hosted on Supabase for management API access.
• For analytics workloads on large tables, consider the ClickHouse MCP instead.

View on the MCPizy directory