Metabase MCP Server
@cuenca-mx
A Model Context Protocol server that enables interaction with Metabase via LLMs
Overview
What is Metabase MCP Server?
An MCP server that exposes Metabase API functionality through the Model Context Protocol interface, enabling AI assistants to interact with Metabase cards, databases, collections, and queries. It is designed for users who want to combine Metabase analytics with AI-driven workflows.
How to use Metabase MCP Server?
Clone the repository and configure your AI assistant client (Claude, Cursor, etc.) with the provided JSON object. Set the command to uv, point directory to the project folder, and supply METABASE_URL and METABASE_API_KEY in the environment variables. The server then exposes tools that your assistant can call.
Key features of Metabase MCP Server
- List Metabase cards
- List available databases
- Execute queries on existing cards
- Run arbitrary queries
- Create new cards
- Create bookmarks and list collections
Use cases of Metabase MCP Server
- Allow an AI assistant to retrieve and present data from Metabase cards.
- Let users run ad‑hoc queries on Metabase databases via natural language.
- Automate the creation of new cards and bookmarks in Metabase.
- Explore and browse collections and database schemas without leaving the chat interface.
FAQ from Metabase MCP Server
What are the runtime requirements?
Python 3.12 or higher and the uv package manager.
How do I configure authentication?
Set the environment variables METABASE_URL (your Metabase instance URL) and METABASE_API_KEY (a valid API key) in the MCP server configuration.
How can I test or debug the server?
Use the MCP Inspector by running make dev, or run unit tests with uv run pytest.
How do I set up the MCP server in my AI client?
Add the provided JSON block to your client’s MCP server settings, replacing the absolute path and environment values with your own.
