Databricks MCP Server

A minimal MCP server that exposes Databricks SQL tools via the Model Context Protocol. Authenticates using OAuth service principal (M2M) — no PAT required.

Tools

Tool	Description
list_tables	List all tables in a given catalog
describe_table	Describe columns of a fully-qualified table
query	Execute a read-only SQL query

Setup

1. Clone and install dependencies:

pip install -r requirements.txt

2. Configure environment variables:

cp .env.example .env
# Fill in your Databricks workspace details and service principal credentials

Required variables:

• DATABRICKS_HOST — Databricks workspace URL
• DATABRICKS_HTTP_PATH — SQL warehouse HTTP path
• DATABRICKS_CLIENT_ID — Service principal client ID
• DATABRICKS_CLIENT_SECRET — Service principal client secret

3. Run the server:

python server.py

The server uses stdio transport by default for MCP communication.

Claude Desktop Configuration

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "databricks": {
      "command": "python",
      "args": ["/absolute/path/to/server.py"]
    }
  }
}

Constraints

• Read-only — write operations (INSERT, UPDATE, DELETE, DROP, etc.) are blocked.
• Partition pruning — queries against Skyscanner tables must include a dt column filter.