This MCP server provides a Databricks integration exposed via the Model Completion Protocol (MCP). It translates MCP tool calls into Databricks REST API requests, allowing LLM-powered workflows to manage clusters, jobs, and notebooks, query workspace files, and run SQL statements against Databricks data. Available tools include: list_clusters to enumerate clusters, create_cluster and terminate_cluster to manage cluster lifecycles, get_cluster and start_cluster for cluster state inquiries and operations, list_jobs and run_job to manage and execute Databricks jobs, list_notebooks and export_notebook to interact with notebooks, list_files to inspect DBFS directories, and execute_sql to run SQL statements. Tools are designed to be invoked from MCP clients and return structured results suitable for downstream reasoning by an AI agent. Async operation is supported, enabling multiple Databricks actions to be performed concurrently where appropriate.

Prerequisites

• Python 3.10 or higher
• uv (recommended MCP runtime manager)

Setup

1. Install uv if you don't have it already:

   MacOS/Linux

   curl -LsSf https://astral.sh/uv/install.sh | sh

   Windows (PowerShell)

   irm https://astral.sh/uv/install.ps1 | iex

   Restart your terminal after installation.

2. Clone the repository:

   git clone https://github.com/JustTryAI/databricks-mcp-server.git cd databricks-mcp-server

3. Set up the project with uv:

   Create and activate virtual environment

   uv venv

   On Windows

   ..venv\Scripts\activate

   On Linux/Mac

   source .venv/bin/activate

   Install dependencies in development mode

   uv pip install -e .

   Install development dependencies

   uv pip install -e ".[dev]"

4. Set up environment variables (example):

   Windows

   set DATABRICKS_HOST=https://your-databricks-instance.azuredatabricks.net set DATABRICKS_TOKEN=your-personal-access-token

   Linux/Mac

   export DATABRICKS_HOST=https://your-databricks-instance.azuredatabricks.net export DATABRICKS_TOKEN=your-personal-access-token

   You can also create an .env file based on the .env.example template.

5. Run the MCP server (via uvx configuration):

   uvx databricks-mcp-server

   The server will start and be ready to accept MCP protocol connections.

Tips and common issues:

• Ensure DATABRICKS_HOST and DATABRICKS_TOKEN are set correctly; missing credentials will result in authentication errors from Databricks APIs.
• If you encounter dependency issues, reinstall in a clean virtual environment and ensure you are using Python 3.10+.
• The MCP server is designed to run asynchronously; for high throughput, consider triggering multiple tool calls concurrently where the MCP client supports it.
• The recommended startup path is using uv with the package name as shown in the configuration; you can also run the included scripts directly for local testing.
• If you need to test specific tools, verify Databricks permissions for the token scope (clusters, jobs, notebooks, workspace, and DBFS as required).
• For local development, refer to the scripts/show_clusters.py and scripts/show_notebooks.py utilities to quickly inspect resources before integrating them into your MCP workflows.