Hey, couldn't find a working Databricks MCP so I forked one and fixed it: https://github.com/markov-kernel/databricks-mcp
This allows you to (using Cursor, Claude Code, Windsurf, Claude, etc.):
* Manage Databricks Clusters: List, create, terminate, start, and get details about your clusters.
* Control Databricks Jobs: List your jobs and run them directly.
* Interact with Notebooks: List notebooks within workspace directories and export them.
* Work with Files: List files and directories in DBFS paths.
* Execute SQL: Run SQL statements directly on your Databricks SQL warehouses or clusters.
It implements the Model Completion Protocol (MCP) to expose these Databricks functionalities as tools for LLM agents.