Overview
Alkemi MCP Server is an STDIO wrapper around the Streamable HTTP MCP Endpoint (https://api.alkemi.cloud/mcp) that connects Alkemi-powered database metadata with MCP clients. It enables querying data sources such as Snowflake, Google BigQuery, and DataBricks and other sources by storing the database metadata, generating appropriate queries, and executing them against the connected databases. This enables teams to share a single MCP Server with consistent query quality across teammates. Access is secured via BEARER_TOKEN, which you obtain as a free API key from datalab.alkemi.ai; optional PRODUCT_ID lets you narrow scope to a single product. The MCP_NAME environment variable allows running multiple MCP Servers in parallel. The README provides configuration examples for Claude Desktop via npx, or for local builds by pointing to a built index.js. Development workflow includes npm install, npm run build, and npm run watch; debugging is supported via the MCP Inspector tool. The project acknowledges the MCP specification and Claude Desktop integration.
Features
Meta-data storage with Alkemi
Stores database metadata using Alkemi to enable accurate query generation and consistent results.
Automatic query generation
Generates proper queries tailored to the connected data sources for MCP clients.
Execution against connected sources
Queries Snowflake, BigQuery, Databricks and other sources via Alkemi backend.
STDIO wrapper for HTTP MCP Endpoint
Provides a STDIO-based interface to the Streamable HTTP MCP Endpoint at api.alkemi.cloud/mcp.
Multi-server support
Supports multiple MCP servers distinguished by MCP_NAME and optional PRODUCT_ID.
Bearer token authentication
Uses BEARER_TOKEN for authentication against the MCP server.
Easy deployment with Claude Desktop integration
Configurable via Claude Desktop JSON config; supports npx and local builds.
Development and debugging tooling
Includes npm build, watch, and MCP Inspector for debugging.
Who Is This For?
- Data teams:Share a standardized MCP server for querying Alkemi-connected data sources across teammates.
- Developers:Integrate into Claude Desktop or other MCP clients using npx or built index.js.
