gnomad

The gnomAD MCP Server exposes a GraphQL-backed interface to query gnomAD data via the MCP protocol. It provides tooling to search for genes, retrieve detailed gene and variant information, and access population genetics data such as allele frequencies and constraint metrics. Typical usage involves starting the server and issuing MCP commands (such as search, get_gene, and get_variant) through your preferred MCP client or CLI integration. This server is intended to enable AI assistants to fetch up-to-date variant data, gene constraints like pLI and LOEUF, and population-specific frequencies directly from the gnomAD dataset.

To use the tools, connect to the running server using the MCP client (or Claude/MCP CLI where configured). Common operations include: searching for genes or variants, retrieving comprehensive gene metadata including constraint scores, and querying exact variant annotations with population frequencies. The available tools are designed to cover gene information, variant analysis, population genetics, transcript data, and more, enabling rich genetic context for AI-assisted reasoning.

Prerequisites:\n- Node.js 18 or higher\n- npm or yarn\n\nInstall from source:\nbash\ngit clone https://github.com/yourusername/gnomad-mcp-server.git\ncd gnomad-mcp-server\nnpm install\nnpm run build\n\nAlternatively, install dependencies and build with yarn:\nbash\nyarn install\nyarn build\n\n\nRun the server (example):\nbash\nnode dist/index.js\n\nEnsure the server starts without errors and is reachable by your MCP client. The build artifacts should be placed under dist and the entry point is index.js as configured in the project.\n\nConfiguration can be adjusted via your MCP CLI or Claude Desktop integration as described in the Configuration section of the README.

Tips and considerations:\n- Ensure you point the MCP server path in your configuration to the actual built entry, e.g. dist/index.js.\n- If you update dependencies or rebuild, reload or restart the MCP server so clients connect to the latest code.\n- The server defaults assume GRCh38 for reference genomes unless overridden by tool parameters.\n- For large queries (e.g., region queries or genome-wide searches), consider rate limits and potential pagination in the MCP client.\n- Environment-specific configuration (like database connections or external GraphQL endpoints) should be handled in the deployment environment; the README focuses on local development.\n- If using Claude Desktop, ensure your Claude config points to the correct local path for the gnomad MCP server.\n- Common issues: missing dist/index.js after build, incorrect node version, or mismatched network ports. Validate by auto-detecting the server startup messages and ensuring the MCP client can connect to the server endpoint.