PostgreSQL Model Context Protocol (PG-MCP) Server

What is PostgreSQL Model Context Protocol (PG-MCP) Server?

PG-MCP is a server implementation of the Model Context Protocol (MCP) for PostgreSQL databases. It provides a comprehensive API for AI agents to discover, connect to, query, and understand PostgreSQL databases through MCP’s resource-oriented architecture. The server includes multi-database support, rich catalog information, built-in extension context (e.g., PostGIS, pgvector), and query explanation tools.

How to use PostgreSQL Model Context Protocol (PG-MCP) Server?

Install via Docker (docker-compose up -d) or manually (uv sync, then python -m server.app). Use the provided test scripts (test.py and example-clients/claude_cli.py) to verify functionality. For AI agents, use the server’s tools (connect, disconnect, pg_query, pg_explain) and explore schema resources via pgmcp://{conn_id}/ URIs.

Key features of PostgreSQL Model Context Protocol (PG-MCP) Server

• Connect tool to register PostgreSQL connection strings and get a secure connection ID
• Read-only SQL execution (pg_query) with connection ID
• pg_explain tool to analyze query execution plans in JSON
• Schema discovery: list schemas, tables, columns, constraints, indexes, extensions
• Sample table data with pagination and approximate row counts
• Built-in YAML-based context for extensions like PostGIS and pgvector

Use cases of PostgreSQL Model Context Protocol (PG-MCP) Server

• AI agents exploring and understanding database schemas autonomously
• Natural language to SQL conversion via Claude-powered CLI
• Analyzing query execution plans for performance tuning
• Programmatic inspection of table relationships, constraints, and indexes

FAQ from PostgreSQL Model Context Protocol (PG-MCP) Server

How do I connect to a database?

Use the connect tool with a PostgreSQL connection string. It returns a secure connection ID used for subsequent operations.

Is the server read-only?

Yes, read-only mode is enforced via transaction settings by default.

What PostgreSQL extensions are supported?

Built-in context is provided for PostGIS and pgvector. Additional extensions can be added via YAML configuration files.

How are database credentials handled?

Credentials are sent once during initial connection and are never exposed in resource URLs—only opaque connection IDs are used.

What runtime and transport does the server require?

It requires Python 3.13+ and uses SSE (Server-Sent Events) transport for production use.