Pgmcp, developed by Subnetmarco, is an MCP server tool built with Go, designed to connect AI assistants with PostgreSQL databases. It enables natural language queries against database content, acting as a crucial bridge for AI agents and developers. Pgmcp supports various installation methods, including pre-compiled binaries and Homebrew, making it accessible across different operating systems. Its core function is to allow AI models to interact with and extract information from PostgreSQL using conversational prompts.

What does Pgmcp connect to?

Pgmcp connects to any existing PostgreSQL database, serving as an MCP (Model Context Protocol) server. It functions as a bridge, allowing AI assistants to interact with these databases using natural language. This connection enables AI agents to formulate queries and retrieve information directly from PostgreSQL. It supports stdio (JSON-RPC) for local client connections, such as when Claude Desktop operates it as a child process. Additionally, Pgmcp supports HTTP (streamable HTTP) for client connections, as indicated by configuration examples.

Core features and tools

Pgmcp exposes several callable tools for database interaction. The ask tool handles natural language questions, converting them into SQL queries with automatic streaming of results. For content discovery, search performs free-text searches across all text columns within the database. The stream tool is designed for advanced streaming of large result sets, incorporating pagination capabilities. Beyond these, LLMs autonomously use list_tables, describe_table, and query to explore database schemas and answer questions effectively.

Who is it for?

Pgmcp is intended for AI assistants and agents, developers, and data analysts who need to interact with PostgreSQL databases using natural language. It provides a means for these groups to query, search, and explore database schemas through conversational interfaces. This tool simplifies data access for those working with AI models and PostgreSQL. Skip this if your primary need is not natural language interaction with PostgreSQL databases via an AI assistant.

How to get started

The server authenticates using an API key passed as an environment variable. Setup involves providing the DATABASE_URL environment variable with the PostgreSQL connection string, and optionally an OPENAI_API_KEY for AI-powered SQL generation.

Use cases

• Querying PostgreSQL databases using natural language to obtain structured SQL results.
• Performing free-text searches across all text columns within the database.
• Enabling LLMs to autonomously explore database schemas and relationships.

Pgmcp provides a specialized solution for connecting AI agents and PostgreSQL databases, enabling natural language interactions. Its support for various client connection methods and a defined set of callable tools makes it adaptable for different development environments. The focus on natural language querying and schema exploration positions it as a valuable asset for developers and data analysts working with AI.