MCP Server Trino

Server for Trino, enabling natural language queries to manage distributed SQL queries, analyze big data across sources, and monitor performance via Claude Desktop, optimized for speed.

Visit Website

Trino MCP Server

This repository provides an MCP (Model-Control-Protocol) server that allows you to list and query tables via Trino using Python.

Overview

MCP: MCP is a protocol for bridging AI models, data, and tools. This example MCP server provides:
- A list of Trino tables as MCP resources
- Ability to read table contents through MCP
- A tool for executing arbitrary SQL queries against Trino
Trino: A fast, distributed SQL query engine for big data analytics. This server makes use of Trino’s Python client (trino.dbapi) to connect to a Trino host, catalog, and schema.

Requirements

Python 3.9+ (or a version compatible with mcp, trino, and asyncio)
trino (the Python driver for Trino)
mcp (the Model-Control-Protocol Python library)

Configuration

The server reads Trino connection details from environment variables:

Variable	Description	Default
`TRINO_HOST`	Trino server hostname or IP	`localhost`
`TRINO_PORT`	Trino server port	`8080`
`TRINO_USER`	Trino user name	required
`TRINO_PASSWORD`	Trino password (optional, depends on your authentication setup)	(empty)
`TRINO_CATALOG`	Default catalog to use (e.g., `hive`, `tpch`, `postgresql`, etc.)	required
`TRINO_SCHEMA`	Default schema to use (e.g., `default`, `public`, etc.)	required

Usage


{
  "mcpServers": {
    "trino": {
      "command": "uv",
      "args": [
        "--directory", 
        "<path_to_mcp_server_trino>",
        "run",
        "mcp_server_trino"
      ],
      "env": {
        "TRINO_HOST": "<host>",
        "TRINO_PORT": "<port>",
        "TRINO_USER": "<user>",
        "TRINO_PASSWORD": "<password>",
        "TRINO_CATALOG": "<catalog>",
        "TRINO_SCHEMA": "<schema>"
      }
    }
  }
}