Overview

PostgreSQL MCP (commonly referred to as Postgres MCP Pro) is an open-source Model Context Protocol (MCP) server designed to give AI agents secure and intelligent access to PostgreSQL databases.

It goes far beyond basic query execution by offering configurable read-only or read/write modes, comprehensive performance diagnostics, automated index recommendations, and production-grade safety features. Built to support the full development lifecycle — from schema exploration and coding to testing, deployment, and ongoing maintenance.

Popular implementations include the original Anthropic reference (now deprecated/archived due to security considerations), community forks, and enhanced versions like Postgres MCP Pro by Crystal DBA.

Features

Configurable Access Modes: Strict read-only transactions (prevents modifications) or controlled read/write with transaction safety.
Schema & Data Exploration: List tables, inspect schemas, columns, indexes, constraints, and run natural language-powered queries.
Performance Analysis: Health checks for buffer cache, vacuum status, replication lag, connection utilization, sequence limits, and more.
Index Tuning & Explain Plans: Advanced index recommendation engine that tests thousands of combinations; support for hypothetical indexes and detailed query explain plans.
Safe SQL Execution: Read-only wrappers, query validation, and safeguards against dangerous operations.
AI-Agent Optimized: Tools designed for low token usage, clear output formatting, and seamless integration with Claude, Cursor, Gemini, and other MCP clients.
Docker & Easy Deployment: Official Docker image for consistent, dependency-free runs.
Extensible: Supports community Postgres, RDS, Aurora, and most PostgreSQL v12+ instances.

Use Cases

AI-Assisted Database Development: Let your AI agent explore schemas, suggest optimal queries, or generate migrations in natural language.
Performance Tuning: Ask "Analyze slow queries" or "Recommend indexes for this table" — receive actionable insights with explain plans.
Production Monitoring: Run health checks, detect bloat, or monitor vacuum/autovacuum without manual SQL.
Safe Data Analysis: Grant read-only access for reporting, analytics, or customer support agents without risking data integrity.
Full Development Workflow: From initial project setup and testing to deployment and ongoing optimization — all driven by conversational AI.

Installation & Quick Start

Using Docker (Recommended)

Pull and run the official image:

docker run -p 8080:8080 crystaldba/postgres-mcp --connection-string "postgresql://user:pass@localhost:5432/mydb"

Python / uv Installation

uv pip install postgres-mcp
uv run postgres-mcp "postgresql://user:password@localhost:5432/dbname"

Configuration for Clients (Claude Desktop, Cursor, etc.)

Add to your MCP config (e.g., ~/.cursor/mcp.json or Claude settings) with the appropriate command/args pointing to the running server.

Full documentation, connection examples, and security best practices are available in the repository.

Security Notes

Prefer read-only mode for untrusted agents.
The original Anthropic reference implementation had a known SQL injection vulnerability (patched in forks and later versions).
Always use connection strings with least-privilege users and consider network-level restrictions.

postgresql mcp