Since ~14:00 UTC on 2026-07-24 the Remote MCP (Rovo) server has been unusable for me — client reports connection timed out after 30000ms, and sometimes HTTP 502. I did direct endpoint testing to isolate it, and this is a server-side hang, not a client/token/config issue:
- Endpoint reachable, token valid. An unauthenticated POST to https://mcp.atlassian.com/v1/mcp returns HTTP 401 in ~0.16s. When the server does answer an authenticated initialize, it returns HTTP 200 — so my OAuth token is valid and not expired.
- Intermittent hang. Authenticated initialize requests hang with 0 bytes received in ~70–80% of attempts (6 of 8 in back-to-back probes), timing out; the rest return 200 in <2s. Occasional HTTP 502 as well.
- Not the endpoint path. Re-testing against https://mcp.atlassian.com/v1/mcp/authv2 shows the same hang rate — so switching the registration URL doesn't help; the fault is in the shared backend behind both paths.
- Ruled out: expired token (it's valid), re-auth (no effect — the token already works when the server responds), and client config (other HTTP MCP servers connect fine from the same client; this is Atlassian-specific).
- Not on the status page — status.atlassian.com shows all-green, so this partial degradation isn't being tracked.
Environment: Claude Code 2.1.218, Streamable HTTP transport, user-scope OAuth, endpoint https://mcp.atlassian.com/v1/mcp.
Questions: Is anyone else seeing intermittent 0-byte hangs / 502s on authenticated MCP requests right now? And can Atlassian confirm/acknowledge a Remote MCP backend degradation? Happy to provide request timestamps or a HAR.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.