Overview
What is CDM MCP Server?
A FastAPI-based service that implements the Model Context Protocol (MCP) to enable AI assistants to interact with Delta Lake tables stored in MinIO through Spark. It is intended for local development or evaluation only and is not approved for production deployment.
How to use CDM MCP Server?
Clone the repository, create required directories, set up a Docker network, and start the services with docker-compose up -d --build. Then access the MCP server at http://localhost:8000/docs. For AI assistant integration, configure an mcp.json file and set up an MCP host.
Key features of CDM MCP Server
- Read-oriented queries against Delta Lake tables via natural language
- Integration with MinIO object storage and Spark
- Implements the Model Context Protocol for AI assistants
- Quick local setup using Docker Compose
- Includes sample data creation and API usage examples
- Supports JupyterHub and Spark Master UI
Use cases of CDM MCP Server
- AI assistants querying Delta Lake tables stored in MinIO
- Local development and evaluation of MCP-based data interactions
- Prototyping natural language data operations with Spark
FAQ from CDM MCP Server
What does CDM MCP Server do?
It allows AI assistants to execute read-oriented queries against Delta Lake tables stored in MinIO through Spark, following the Model Context Protocol.
Is CDM MCP Server approved for production?
No. The README explicitly states it is not approved for deployment to any production environment, including CI, without explicit approval from KBase leadership.
What are the runtime requirements?
Docker, Docker Compose, and a local network (cdm-jupyterhub-network). The service runs locally via docker-compose up.
Where does data live?
Data is stored in MinIO (object storage) and accessed via Spark. The service creates a local shared workspace directory (cdr/cdm/jupyter/cdm_shared_workspace).
What transport does the server use?
The server exposes a REST API at http://localhost:8000/docs and implements the Model Context Protocol for AI assistant integration.
