Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Databricks MCP Server list tables in the sales catalog"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
A Python MCP server to interact with Databricks — run SQL queries, manage clusters, trigger jobs, run notebooks, explore Unity Catalog metadata, manage secrets, and interact with DBFS.
pip install -r requirements.txtEdit the .env file:
DATABRICKS_HOST=https://<your-workspace>.azuredatabricks.net
DATABRICKS_TOKEN=<your-access-token>
DATABRICKS_SQL_HTTP_PATH=/sql/1.0/warehouses/<your-warehouse-id>
DATABRICKS_WAREHOUSE_ID=<your-warehouse-id>python -m databricks_mcp.serverRelated MCP server: Databricks MCP Server
Tool | Description |
| Run a SQL query (auto-wakes warehouse) |
| Show execution plan of a SQL query |
Tool | Description |
| List all clusters and their states |
| Get current state of a specific cluster |
| Start a cluster |
| Stop a cluster |
| Restart a running cluster |
| Scale cluster workers up/down |
| Poll until cluster is RUNNING |
| Get recent events/errors for a cluster |
Tool | Description |
| List all SQL Warehouses |
| Get detailed status of a warehouse |
| Start a SQL Warehouse |
| Stop a SQL Warehouse |
| Change warehouse size |
Tool | Description |
| List all Databricks jobs |
| Create a new job with a notebook task |
| Update a job's name, notebook, or cluster |
| Delete a job |
| Trigger a job run |
| Get status of a job run |
| Cancel a running job run |
| List recent runs of a job |
Tool | Description |
| List notebooks in a workspace path |
| List all objects in a workspace path |
| Create a folder in workspace |
| Run a notebook (fire and forget) |
| Run a notebook and wait for output |
| Export a notebook (HTML/SOURCE/JUPYTER/DBC) |
| Import/upload a notebook to workspace |
| Delete a notebook or folder |
Tool | Description |
| List all catalogs |
| List schemas in a catalog |
| List tables in a catalog.schema |
| Get column schema of a table |
| Preview top N rows of a table |
| Create a Delta table from DDL |
| Drop a table |
| Run OPTIMIZE (+ optional ZORDER) |
| Run VACUUM to clean up old files |
Tool | Description |
| List all secret scopes |
| List secret keys in a scope (no values) |
Tool | Description |
| List files in a DBFS path |
| Delete a file or folder from DBFS |
| Upload a file to DBFS (base64 content) |
| Create a directory in DBFS |
Add this to your claude_desktop_config.json:
{
"mcpServers": {
"databricks": {
"command": "python",
"args": ["-m", "databricks_mcp.server"],
"cwd": "/path/to/MCP"
}
}
}This server cannot be deployed
Your Databricks Lakehouse in natural language: run SQL on your SQL warehouses, track long-running qu
Remote data science agents for Snowflake, Databricks & BigQuery in Claude/Cursor via MCP
The grounded data layer for any LLM: governed SQL, metrics, lineage and catalog over your data.
List datasets, schemas, run APL queries, and use prompts for exploration, anomalies, and monitoring.
