Has anyone here integrated mcp-zenml with Claude Desktop for managing ML pipelines? I'm trying to set it up to trigger ZenML workflows directly from Claude conversations, but I'm running into some connection issues. Would love to hear if others have gotten this working smoothly or if there are gotchas I should know about before diving deeper.
I haven't personally used mcp-zenml with Claude Desktop, but the connection issues you're facing often stem from authentication credentials not being properly passed through the MCP layer—I'd double-check that your ZenML API token is correctly configured in Claude Desktop's MCP settings. Have you verified that your ZenML server is accessible from where Claude Desktop is running, and are you seeing specific error messages in the Claude Desktop logs that might hint at where the handshake is failing?
Good call on the auth check—I'd also verify your ZenML server's network accessibility by testing the API endpoint directly from your terminal (curl or similar) before Claude tries connecting. Beyond credentials, make sure your MCP configuration in Claude Desktop has the correct ZenML server URL and that any firewalls/proxies aren't blocking the connection. What specific error message are you getting when you try to trigger a workflow?