How to Install Databricks MCP in Claude Code
Official Databricks integration — run notebooks, manage clusters, and query lakehouse data through AI
What Databricks MCP does
The official Databricks MCP server gives AI agents access to lakehouse analytics and data engineering workflows. It can run notebooks, inspect clusters, and query Delta tables. Features: - Execute notebooks and review job results - Query Delta Lake and Unity Catalog tables - Manage clusters and compute resources - Inspect data lineage and table schemas - Run SQL analytics and generate dashboards - Access MLflow model registry and experiments Databricks MCP is built for data teams who want AI to take action on analytics, ML, and lakehouse assets.
Installation steps
- 1Copy the config below
- 2Open or create .claude/mcp.json in your project root
- 3Paste inside "mcpServers": {}
- 4Reload Claude Code
Configuration
File location: .claude/mcp.json (project root) or ~/.claude/mcp.json (global)
{
"databricksMcp": {
"command": "npx",
"args": [
"-y",
"@databricks/mcp"
],
"env": {
"DATABRICKS_HOST": "https://adb-...",
"DATABRICKS_TOKEN": "..."
}
}
}Install in other clients
For maintainers
If you maintain Databricks MCP, add this badge to your README to show it's verified on CuratedMCP:
[](https://curatedmcp.com/marketplace/databricks-mcp)