- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
a month ago
Hi @Bihaag_N,
Thanks for checking. If the same remote Power BI MCP server works in both Genie Code and AI Playground, that strongly suggests the MCP server itself is properly configured, rather than a basic transport or auth issue with the remote server. Your latest note also adds that the same connection is not being picked up when you try to use it as context in a Supervisor Agent.
That lines up with the public docs, which say external MCP servers can be validated in AI Playground, used in Genie Code, and used programmatically in agents through the Databricks-managed proxy. See Install an external MCP server, Connect Genie Code to MCP servers, and Use external MCP servers in agents.
The Supervisor Agent docs also explicitly list the External MCP server as a supported tool / sub-agent type, with the key requirement being USE CONNECTION on the Unity Catalog connection. So if that permission is already in place and the connection still does not appear there, I would lean toward this being a surface-specific product issue in Genie UI/chat/agent selection, rather than something inherently unsupported about your remote MCP server.
So based on what you’ve shared, I think..
- The remote MCP server itself appears valid because it works in Genie Code and AI Playground
- The problem seems isolated to Genie UI and possibly the Supervisor Agent connection picker
- If permissions are correct, this looks more like a Databricks UI / product inconsistency than a connection-side fix for the Power BI MCP server
I’d recommend raising a support ticket and including..
- Confirmation that the same remote MCP works in Genie Code
- Confirmation that it also works in AI Playground
- Che exact TypeError: unhashable type: 'list' from Genie UI
- The fact that the connection also does not get picked up when added as Supervisor Agent context
If this answer resolves your question, could you mark it as “Accept as Solution”? That helps other users quickly find the correct fix.
Ashwin | Delivery Solution Architect @ Databricks
Helping you build and scale the Data Intelligence Platform.
***Opinions are my own***