🚀 MCP Databricks

A powerful Databricks integration for AI assistants via Model Context Protocol

📖 Introduction

MCP Databricks seamlessly connects AI assistants to your Databricks workspaces through the Model Context Protocol (MCP). Built with Python, it provides a rich collection of tools for managing virtually every aspect of your Databricks environment.

With this server, AI assistants like Claude can:

• 🔧 Manage compute resources with precision
• 📊 Execute SQL queries and analyze results
• 📁 Organize and manipulate workspace objects
• ✨ And much more!

🔍 Prerequisites

• 🐍 Python 3.11 or higher
• 💻 A Databricks workspace
• 🔑 Databricks Personal Access Token (PAT)
• 📦 Required Python packages (installed in setup)

🚀 Quickstart

1️⃣ Clone the repository

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

2️⃣ Configure environment variables

Create a .env file in the project root with your Databricks credentials:

# Databricks API Configuration
DATABRICKS_HOST="https://adb-<your workspace uri>.azuredatabricks.net/"
DATABRICKS_TOKEN="dapi_<your_token_here>"

# Server Configuration
SERVER_HOST="0.0.0.0"
SERVER_PORT="8000"
DEBUG="false"
TRANSPORT="stdio"

# Logging Configuration
LOG_LEVEL="INFO"

💡 Tip: For further instructions, You can use the env.sample file as a template.

3️⃣ Choose Your Installation Method

You can use MCP Databricks in two ways:

Option A: Docker (Recommended for production)

1. Build the Docker image:

docker build -t databricks-mcp .

2. Configure in Cursor with the following mcp.json entry:

{
  "databricks-mcp-docker": {
    "command": "docker",
    "args": [
      "run",
      "--rm",
      "-i",
      "--name", "databricks-mcp",
      "--env-file", "<path/to/.env>",
      "databricks-mcp"
    ]
  }
}

💡 Note: Replace <path/to/.env> with the absolute path to your .env file.

Option B: Local Installation with uv

1. Install uv (if not already installed):

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

💡 Note: See uv installation documentation for alternative installation methods.

2. Set up virtual environment:

uv venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

3. Install dependencies with uv:

uv sync

4. Configure in Cursor with the following mcp.json entry:

{
  "databricks-mcp-stdio": {
    "command": "uv",
    "args": [
      "--directory",
      "<repository directory>",
      "run",
      "main.py"
    ]
  }
}

💡 Note: Replace <repository directory> with the absolute path to your cloned repository.

🚀 Usage

Running the MCP Server

If not using Docker or Cursor integration, start the server with:

python main.py

or

uv run main.py

Connecting to Claude or other MCP clients

This server uses the stdio transport for seamless compatibility with Claude Desktop and other MCP clients. After installing the server, you can immediately connect to it using your preferred MCP client.

🧰 Tools and Capabilities

The MCP Databricks server provides a comprehensive toolkit for managing your Databricks environment:

💻 Cluster Management

Tool	Description
list_clusters	List all Databricks clusters in the workspace
create_cluster	Create a new Databricks cluster with customizable settings
delete_cluster	Delete a Databricks cluster by ID
start_cluster	Start a terminated Databricks cluster
list_node_types	List all available node types for Databricks clusters
list_spark_versions	List all available Spark versions for Databricks clusters
get_cluster	Get detailed information about a specific Databricks cluster

📦 Library Management

Tool	Description
install_libraries	Install libraries (JAR, WHL, PyPI, Maven, CRAN, etc.) on a running cluster

🖥️ Command Execution

Tool	Description
execute_command	Execute a command (Python, Scala, SQL) on a running Databricks cluster
create_execution_context	Create an execution context for interactive command sessions

📊 SQL Warehouse Management

Tool	Description
list_sql_warehouses	List all SQL warehouses in the workspace
create_sql_warehouse	Create a new SQL warehouse with configurable size and settings

📁 Workspace Objects

Tool	Description
delete_workspace_object	Delete an object from the Databricks workspace
get_workspace_object_status	Get the status of an object in the Databricks workspace
import_workspace_object	Import an object (notebook, file, etc.) into the workspace
create_workspace_directory	Create a directory in the Databricks workspace

---