databricks-mcp-server-by-cdata

What is databricks-mcp-server-by-cdata?

A read-only Model Context Protocol (MCP) server that lets LLMs like Claude Desktop query live Databricks data using natural language. It wraps the CData JDBC Driver for Databricks and exposes Databricks tables and columns as MCP tools.

How to use databricks-mcp-server-by-cdata?

Clone the repository, build with Maven (mvn clean install), download and license the CData JDBC Driver for Databricks, create a .prp file with your connection details, then add the server to your Claude Desktop configuration. The server runs via stdio and can only be used with clients on the same machine.

Key features of databricks-mcp-server-by-cdata

• Read-only access to Databricks data via natural language queries
• Exposes Databricks tables and columns as MCP tools
• Built on the CData JDBC Driver for Databricks
• SQL SELECT queries executed automatically by the LLM
• Simple setup with a .prp configuration file
• MIT licensed and open-source

Use cases of databricks-mcp-server-by-cdata

• Ask natural language questions about Databricks data (e.g., "What is the correlation between my closed won opportunities and account industry?")
• Retrieve live Databricks data during LLM conversations without writing SQL
• Enable Claude Desktop to explore and summarize Databricks tables

FAQ from databricks-mcp-server-by-cdata

Is this server read-only?

Yes. The README explicitly states it builds a read-only MCP server. For full CRUD capabilities, CData offers a separate beta server.

What are the runtime requirements?

You need Java, Maven (to build), and the CData JDBC Driver for Databricks (trial or licensed). The server runs on Windows, Mac, or Linux.

How does the server connect to Databricks?

It uses a JDBC connection string configured in a .prp file. The CData driver handles OAuth authentication when needed.

What transport protocol does it use?

The server uses stdio transport, so it must run on the same machine as the MCP client (e.g., Claude Desktop).

Does it support write, update, or delete operations?

No. Only read operations (via get_tables, get_columns, and run_query for SELECT queries) are provided in this version.