metabase

This MCP server integrates Metabase with the Model Context Protocol to provide smart data access tooling. It exposes a set of unified commands—list, retrieve, search, execute, and export—to interact with Metabase resources such as cards, dashboards, tables, databases, and collections. Authentication can be performed via API keys or user credentials, and the server runs with Read-Only mode enabled by default to restrict write operations. Use the export feature to generate large datasets in CSV, JSON, or XLSX formats, with caching and configurable timeouts to optimize performance during repeated queries. The tools are designed to support both ad-hoc queries and saved card executions, while providing safe, structured responses for downstream AI workflows.

Prerequisites:\n- Node.js 18.0.0 or higher (per the project requirements)\n- Git\n- Access to an Metabase instance for integration\n\nInstallation steps:\n1) Clone the repository:\n``` git clone https://github.com/jerichosequitin/metabase-mcp.git cd metabase-mcp

2) Install dependencies and build the project:\n```
npm install
npm run build

3. Run locally or configure via MCP client:\n- Local development using Node.js: replace the example with the built path:\n``` { "mcpServers": { "metabase-mcp": { "command": "node", "args": ["/path/to/metabase-mcp/build/src/index.js"], "env": { /* see Manual Configuration for options */ } } } }

4) Optional: use npx or Docker as described in the README for deployment in environments without local Node.js installation. See the manual configuration example in the repository for the exact environment variables and flags.

Tips & caveats:\n- Default Read-Only mode is METABASE_READ_ONLY_MODE=true; change to false only if you need write access (be mindful of security implications).\n- API key authentication is recommended for production usage; you may also use email/password as an alternative.\n- The EXPORT_DIRECTORY controls where large export files are saved; ensure the path is accessible to the running process.\n- CACHE_TTL_MS and REQUEST_TIMEOUT_MS help tune performance and reliability for slow or large Metabase instances.\n- When using Docker, be sure to pass environment variables with -e and mount volumes for exports if needed.\n- If you encounter token or authentication errors, verify that METABASE_URL is reachable from the MCP host and that credentials have the required permissions.\n- The build outputs and MCP package are designed to work with the MCP client configuration in JSON, so keep your client config aligned with the mcp_config example.