Our users reported that Gmail and Google Calendar integrations were failing for them on our product. To investigate, I tried connecting to the MCP server directly through Postman and the connection was hanging. MCP URL: https://apollo.composio.dev/v3/mcp/49b915c1-f505-45cb-a833-8b4a5e672490/mcp?include_composio_helper_actions=false&user_id=db5e68da-0d4a-4d5d-b0d5-a82d138f8c7c After some time, it recovered. However some critical actions for our users were not correctly completed. I'd like to know if there's any root cause for the issue and if it's expected to happen more in the future as the robustness of our integrations is crucial to the correct functioning of the product

Hi there, Sorry for the degraded performance your users hit. I want to pin down the real cause rather than guess. On the Postman check: these MCP endpoints are POST-based (JSON-RPC over streamable HTTP), and a valid POST opens a response stream that stays open by design, so a client like Postman can look like it is "hanging" even when the request is healthy. (A plain GET is a separate case, it is rejected immediately with a 405, so it would not hang.) Either way, the Postman behavior on its own does not confirm an outage. The actions your users actually ran are the better signal. Checking our side for those two server IDs around recent activity, we are seeing requests come in and get processed without server-side errors or timeouts, and I could not find traffic tied to the

user_id

in the URL you shared (

db5e68da-...

). That is exactly why I need a few specifics to trace the real failures: • the exact

user_id

and/or connected account that failed (the one in the URL may not be the one your users were on) • approximate timestamps with timezone for the Gmail/Calendar and Outlook failures • the

x-request-id

response header from any failed or hanging request, if you captured it • whether it failed at the connect step or during a specific tool call, and the HTTP status/body if you have it We are tracking intermittent slowness on some Google MCP connections internally, and I do not have a timeline to share yet. Send those identifiers and we will trace exactly what happened on your accounts. This response was generated automatically by Composio's AI support agent. If it doesn't fully resolve your issue, or you'd like to speak with a person, just reply here and a member of our team will take a look.