Databricks MCP Server

What is Databricks MCP Server?

The Databricks MCP Server is a Model Context Protocol (MCP) server that connects to the Databricks API, enabling LLMs to run SQL queries on Databricks SQL warehouses, list jobs, and retrieve job status. It is intended for developers who want to let AI assistants interact with their Databricks workspace.

How to use Databricks MCP Server?

Set up the server by cloning the repository, creating a virtual environment, installing dependencies, and configuring a .env file with your Databricks host, personal access token, and SQL warehouse HTTP path. Start the server with python main.py. Optionally test the connection with python test_connection.py or use the MCP inspector via npx @modelcontextprotocol/inspector python3 main.py.

Key features of Databricks MCP Server

• Run SQL queries on Databricks SQL warehouses
• List all Databricks jobs in your workspace
• Get status of a specific Databricks job
• Get detailed information about a Databricks job
• Integrates with LLMs through the Model Context Protocol

Use cases of Databricks MCP Server

• Ask an LLM to show all tables in the database
• Instruct an LLM to count records in a specific table
• Have an LLM list all your Databricks jobs
• Request an LLM to check the status of job #123
• Ask an LLM to show details about a particular job

FAQ from Databricks MCP Server

What prerequisites are needed to run Databricks MCP Server?

Python 3.7+ and a Databricks workspace with a personal access token, a SQL warehouse endpoint, and permissions to run queries and access jobs.

How do I obtain my Databricks credentials?

Your Databricks host is the instance URL (e.g., your-instance.cloud.databricks.com). Create a personal access token under User Settings > Developer > Access tokens. For the HTTP path, go to SQL Warehouses, select your warehouse, and copy the connection details.

How do I store credentials securely?

Place them in a .env file containing DATABRICKS_HOST, DATABRICKS_TOKEN, and DATABRICKS_HTTP_PATH. Never commit this file to version control.

What MCP tools does the server provide?

Four tools: run_sql_query (executes SQL), list_jobs (lists all jobs), get_job_status (status by job ID), and get_job_details (detailed info by job ID).

What should I do if connection fails?

Verify the Databricks host does not include https://, ensure your SQL warehouse is running, check the token’s permissions, and run the included test script python test_connection.py.