Databricks SQL
The Databricks SQL MCP server is Databricks' managed HTTP server that lets agents run AI-generated SQL against your Unity Catalog tables. Queries run over Streamable HTTP with access governed by Unity Catalog permissions.
Which Databricks MCP do you want?
| If you need | Use |
|---|---|
| Run SQL queries, validate syntax, or author data pipelines against Unity Catalog | Databricks SQL (this page) |
| Ask a business question in natural language and get an answer grounded in your semantic layer | Databricks Genie One MCP |
| Unity Catalog metadata, data discovery, and lineage via a local server | Databricks (by Willow) |
Setting up
-
Find your workspace hostname. In the Databricks workspace, the hostname is the part of the URL after
https://, for exampleadb-1234567890123456.7.azuredatabricks.net. -
Create a Databricks OAuth application. In the Databricks account console, go to Settings → App Connections → Add connection:
- Name: choose any name, e.g.
willow-mcp - Redirect URLs: paste in the Willow redirect URL shown in the setup wizard
- Client type: uncheck Generate a client secret for a public client, or leave it checked for a confidential client
- Scopes: add
sql
- Name: choose any name, e.g.
-
In Willow, go to MCP Servers → + Add MCP server, find the Databricks SQL card with the Official badge, and click + Use.
-
Enter your workspace hostname (replacing
YOUR_WORKSPACE_HOSTNAMEin the server URL), then paste the Client ID from the Databricks integration record (and Client Secret if you created a confidential client).
Databricks does not support dynamic client registration, so you must create the OAuth app yourself before connecting.