Description When generating an MCP URL via the Composio SDK using a valid API key and server ID, the API returns a 401 Unauthorized error. The actual root cause is an internal Redis Cluster slot allocation error, not an authentication failure. Request Details - Server ID: b55dbc24-9e19-42f9-8906-8dd999d57dd0 - Service: google_calendar - Operation: composio.mcp.generate(userId, serverId, { connectedAccountIds }) Error Response { "error": { "message": "Redis error: All keys in the pipeline should belong to the same slots allocation group", "code": 10401, "slug": "HTTP_Unauthorized", "status": 401, "request_id": "ce95f922-9185-4a6a-b2d7-e375525e3367", "suggested_fix": "Please check you are using a valid API key." } } Issues 1. Incorrect HTTP status code — A Redis infrastructure error is being returned as 401 Unauthorized. This should be a 500 Internal Server Error or 503 Service Unavailable. 2. Misleading error message — "Please check you are using a valid API key" sends developers on a wild goose chase when the API key is valid. The actual problem is server-side Redis key routing. 3. Redis Cluster misconfiguration — The error "All keys in the pipeline should belong to the same slots allocation group" indicates that a Redis pipeline operation is targeting keys that hash to different cluster slots. This typically means multi-key operations need {hash_tag} prefixes to co-locate keys on the same shard. Expected Behavior - Internal infrastructure errors should return 5xx status codes, not 401 - The error message should not suggest API key issues when the root cause is server-side - Redis pipeline operations should use hash tags to ensure slot co-location Impact - Intermittent failures on MCP URL generation for connected services - Breaks agent tool loading at chat time for affected users - Misleading error makes debugging difficult on the integrator side

Hey @JoelCodes, we are checking this. Also, could you specify if this is always reproducible?

Don’t easily reproducible, it’s a downstream error from your redis it seems.

Got it, I've escalated the issue with the team and will revert back with any updates.