# DatabricksMCP

An open-source MCP server that lets AI agents query Databricks SQL, browse Unity Catalog, run jobs and use vector search, under per-tool policy controls.

DatabricksMCP is a Model Context Protocol server that exposes governed access to a Databricks environment's data, SQL, workflows, compute, workspace assets, governance, ML and AI features, and cost information, so AI assistants can inspect and operate the platform without bypassing its security or operational controls.

It currently exposes 63 read tools across 11 capability packs, including Unity Catalog, SQL, Jobs, Lakeflow pipelines, compute, governance, workspace assets, MLflow, and AI features such as Model Serving, Vector Search and Genie, plus a first gated write pack. The read surface is always available; state-changing tools are disabled by default, and when enabled are gated per tool with destructive actions requiring an approval token and full auditing.

It uses Databricks' own unified authentication rather than accepting credentials as model-generated arguments, and bounds list sizes, SQL result sizes, byte limits and wait times by default. It is an independent open-source project, not affiliated with or endorsed by Databricks, Inc.

Website: https://github.com/Purvanshsingh/DatabricksMCP