databricks-mcp

27
cloud_platformsdeveloper_toolsdatabasesai_chatbotos_automation

Markov Databricks MCP is a Model Completion Protocol (MCP) server that allows interactions with Databricks services through the MCP protocol, enabling integration with LLM-powered tools. It features comprehensive tools for managing Databricks clusters, jobs, and notebooks, with support for async operations.

Markov Databricks MCP

A Model Completion Protocol (MCP) server for Databricks that facilitates interaction with Databricks clusters, jobs, and notebooks using LLM-powered tools.

Features

  • Implements MCP protocol for Databricks interaction
  • Integrates with Databricks REST API
  • Exposes Databricks features as MCP tools
  • Supports asynchronous operations with asyncio

Available Tools

  • Listing, creating, starting, and terminating clusters
  • Listing and running jobs
  • Listing and exporting notebooks
  • Executing SQL statements

Installation

  • Requires Python 3.10+
  • Use uv package manager for setup

Running the MCP Server

  • Start the server for testing or development

Integrating with AI Clients

  • Register the server with clients like Cursor or Claude CLI

Querying Databricks Resources

  • Utility scripts available for resource viewing

Project Structure

  • Contains source, test, script, example, and doc directories

Development

  • Follows PEP 8 code standards
  • Utilizes various linting tools

Testing

  • Uses pytest for testing with coverage goal of 80%

Documentation

  • API documentation generated with Sphinx

Examples

  • Example usage in examples/ directory

Contributing

  • Contributions welcome with adherence to coding standards

License

  • Licensed under the MIT License.