Custom MCP Server for Databricks

Custom Model Context Protocol (MCP) server with OBO (On-Behalf-Of) authentication, designed for deployment as a Databricks app.

Features

The server implements three MCP tools:

• test_connection: Test server connectivity and OBO authentication status

• echo: Echo back a message (useful for testing basic functionality)

• get_user_info: Retrieve current user information via OBO authentication

Related MCP server: Databricks MCP Server

Authentication

This server uses Databricks OBO (On-Behalf-Of) authentication with ModelServingUserCredentials, allowing it to act on behalf of the authenticated user making requests.

Deployment to Databricks

Prerequisites

1. Install Databricks CLI

2. Authenticate to your workspace:

   databricks auth login --host https://<your-workspace-hostname>

Deploy Steps

1. Create the Databricks app:

   databricks apps create mcp-cust-kat

2. Upload and deploy:

   DATABRICKS_USERNAME=$(databricks current-user me | jq -r .userName)
   databricks sync . "/Users/$DATABRICKS_USERNAME/mcp-cust-kat"
   databricks apps deploy mcp-cust-kat --source-code-path "/Workspace/Users/$DATABRICKS_USERNAME/mcp-cust-kat"

3. Access your app:

   • The app URL will be available in the Databricks UI

   • MCP endpoint: https://<app-url>/mcp

   • Health check: https://<app-url>/health

API Endpoints

• GET / - Root endpoint with server information

• GET /health - Health check endpoint with OBO authentication status

• POST /mcp - Main MCP JSON-RPC endpoint

Testing the Server

Once deployed, you can test the connection using the available MCP tools:

1. test_connection: Verifies the server is running and shows OBO authentication status

2. echo: Simple echo test to verify request/response functionality

3. get_user_info: Returns detailed information about the authenticated user

Local Development

Run locally for testing:

uv run python server.py

The server will start on port 8000 with both HTTP and MCP protocol support.

Project Structure

• server.py - Main FastAPI application with MCP protocol implementation

• app.yaml - Databricks app configuration

• pyproject.toml - Python project configuration and dependencies

• requirements.txt - Direct dependencies for the application