Claude Desktop connects to Databricks through the CData MCP Server for Databricks using the Model Context Protocol (MCP).

When you launch Claude, it automatically detects MCP connections registered during setup. These connections are created and saved using the MCP Configuration Tool, which is installed with the MCP Server.

Each saved connection is recognized by Claude and can be referenced by name when asking questions.

Installing Claude

Claude Desktop is a standalone application that enables natural language querying across supported data sources using the Model Context Protocol (MCP).

To connect to Databricks via MCP, you first need to install the Claude Desktop application.

You can download the latest version from the official Claude Download Page.

After installing Claude Desktop, you can configure your MCP connection using the CData MCP Driver for Databricks.

How Claude Uses the Server

When you ask Claude a question, it launches the corresponding MCP Server connection, reads the available tool descriptions provided by the connector, and generates a secure query in the background.

There is no need to write SQL or teach Claude how to query your data. Each tool is self-describing, and Claude consumes those descriptions to determine how to retrieve the right information.

Example Prompts

After setup, you can ask Claude natural-language questions like:

• "What are the total sales by product category this quarter?"
• "Which customers placed more than 5 orders last month?"
• "Show average transaction size by region for the last 6 months."
• "List all jobs that failed in the past 7 days."
• "Which tables in the 'analytics' schema were updated most recently?"
• "Summarize compute usage by cluster name for this week."
• "What are the top-selling products by revenue in Q1?"
• "List the number of rows in each table in the 'default' database."
• "Show the daily active users over the last 30 days."
• "Which notebooks were run most often this month?"

Claude securely executes the query using the MCP Server and responds using the live results from Databricks.