PostgreSQL MCP Server (Model Context Protocol)

What is PostgreSQL MCP Server?

A FastMCP-powered Model Context Protocol server that gives AI assistants (Claude, Cursor, etc.) full, structured access to a PostgreSQL database — no SQL knowledge required from the user.

How to use PostgreSQL MCP Server?

Clone the repository, create a virtual environment, install dependencies from requirements.txt, and configure a .env file with database credentials. Then connect an MCP-compatible client (e.g., Claude Desktop) by adding the provided JSON block to its MCP config file. Alternatively, run manually with python postgres_mcp_server.py.

Key features of PostgreSQL MCP Server

• Exposes database as MCP Resources (read-only) and MCP Tools (read/write)
• Parameterised queries throughout — no SQL injection
• Structured stdout logs on every tool/resource call
• Schema inspection: list tables, column definitions
• Read/write/DDL operations via tools
• Raw SQL execution via the execute_query tool

Use cases of PostgreSQL MCP Server

• AI assistant querying database in natural language
• Automated schema exploration and data preview
• Performing insert, update, delete operations through conversational interface
• Creating and dropping tables via AI assistant

FAQ from PostgreSQL MCP Server

What transport does PostgreSQL MCP Server use?

The server uses stdio transport (MCP protocol over standard input/output).

What are the runtime requirements?

Python 3.8+ and a PostgreSQL server (local or remote) with the uv package runner (recommended) or pip.

How are database credentials configured?

Via a .env file with POSTGRES_HOST, POSTGRES_PORT, POSTGRES_DB, POSTGRES_USER, and POSTGRES_PASSWORD, or directly in the client config's env block.

What dependencies are required?

mcp[cli], asyncpg, psycopg2-binary, python-dotenv, httpx, uvicorn, and uv.

What are the limits on data access?

The postgres://data/{table_name} resource returns only the first 100 rows as a safe preview. Other tools allow configurable limits.