Databricks MCP Server

MCP-practice

this is repository for all my MCP code practices

Databricks MCP Server

A Model Context Protocol (MCP) server that provides tools for interacting with Databricks SQL Analytics.

Files Overview

• venv/test_mcp_cloud.py: Main MCP server script providing Databricks SQL tools
• venv/databricks_connection.py: Connection test script to verify Databricks credentials
• venv/run_inspector.ps1: PowerShell script to launch MCP inspector with environment variables
• venv/.env: Environment variables file (create this with your credentials)
• README.md: This documentation file

Features

• Query Databricks: Execute SELECT queries on Databricks tables
• Update Databricks: Perform INSERT, UPDATE, DELETE operations
• List Tables: Browse available tables in your schema
• Inspect Schema: Get column information for specific tables

Prerequisites

• Python 3.8+
• Databricks workspace with SQL Analytics enabled
• Personal Access Token with appropriate permissions

Installation

1. Clone this repository
2. Create a virtual environment:

   python -m venv venv
   venv\Scripts\activate  # On Windows

1. Install dependencies:

   pip install fastmcp databricks-sql-connector sqlalchemy python-dotenv

Configuration

1. Create a .env file in the venv/ directory:

   DATABRICKS_SERVER_HOSTNAME=your-workspace.databricks.com
   DATABRICKS_HTTP_PATH=/sql/1.0/warehouses/your-warehouse-id
   DATABRICKS_TOKEN=your-personal-access-token
   DATABRICKS_CATALOG=your-catalog-name
   DATABRICKS_SCHEMA=your-schema-name

1. Test your connection:

   python venv/databricks_connection.py

Usage

Testing the MCP Server

Using the PowerShell script (recommended):

1. Edit venv/run_inspector.ps1 and update the paths and your Databricks credentials
2. Run: ./venv/run_inspector.ps1

Manual command: ``bash npx @modelcontextprotocol/inspector venv/Scripts/python.exe venv/test_mcp_cloud.py ``

Important: Before running run_inspector.ps1, you must edit the file and update:

• The $PYTHON_EXE and $SERVER_PY paths to match your system
• All the credential variables ($TOKEN, $HOSTNAME, $PATH, $CATALOG, $SCHEMA) with your actual Databricks information

Claude Desktop Integration

To use this MCP server with Claude Desktop:

1. Locate the config file:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

1. Add the server configuration:

   {
     "mcpServers": {
       "databricks": {
         "command": "python",
         "args": ["path/to/your/weather-mcp-server/venv/test_mcp_cloud.py"],
         "env": {
           "DATABRICKS_SERVER_HOSTNAME": "your-workspace.databricks.com",
           "DATABRICKS_HTTP_PATH": "/sql/1.0/warehouses/your-warehouse-id",
           "DATABRICKS_TOKEN": "your-personal-access-token",
           "DATABRICKS_CATALOG": "your-catalog-name",
           "DATABRICKS_SCHEMA": "your-schema-name"
         }
       }
     }
   }

1. Restart Claude Desktop to load the new MCP server

1. Test in Claude: Ask Claude to "list my Databricks tables" or "query my data"

Using with MCP Clients

The server provides these tools:

• query_databricks(sql_query): Execute SELECT queries (auto-limits to 100 rows)
• update_databricks(sql_command): Execute DML operations
• list_cloud_tables(limit): List available tables
• inspect_cloud_schema(table_name): Get table schema information

Example Usage

# Query data
result = query_databricks("SELECT * FROM customers WHERE region = 'US'")

# Update data
result = update_databricks("UPDATE customers SET status = 'active' WHERE id = 123")

# List tables
tables = list_cloud_tables(10)

# Get schema
schema = inspect_cloud_schema("customers")

Security Notes

• Only SELECT queries are allowed in query_databricks
• Only INSERT, UPDATE, DELETE are allowed in update_databricks
• Table names are validated to prevent SQL injection
• Results are truncated to prevent large payloads
• Queries are automatically limited to 100 rows unless specified

Troubleshooting

1. Connection fails: Check your .env file and token permissions
2. No tables found: Verify catalog and schema names
3. Inspector doesn't connect: Ensure correct file paths in commands
4. Large result errors: The server automatically limits/truncates results

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Test with the inspector
5. Submit a pull request