video1.0<iframe src="https://www.loom.com/embed/a3b712f4442f40c6a01dd923d26c1a11" frameborder="0" width="1662" height="1246" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>12461662Loomhttps://www.loom.com12461662https://cdn.loom.com/sessions/thumbnails/a3b712f4442f40c6a01dd923d26c1a11-5e19c39634028b52.gif198.513I show an example of connecting Databricks VectorSearch to Thunk. In Databricks, I use a customer review dataset and a VectorSearchIndex with full text and embedding search, then copy the MCP server URL from the compute VectorSearch page. In Thunk, I add a custom connection named Review Search, paste the MCP server URL, and enter my Databricks personal access token as a bearer authorization header. After adding it, I test the tool by searching for oatmeal and confirm results return with vector search scores. I then enable the tool in my workflow and update the instructions to use it for related review search.