Hey, thanks for the report. We need more data for diagnostics.
Could you please:
Attach console logs from Help > Toggle Developer Tools > Console at the moment the state changes
Send the Request IDs from chat for the message where the agent created the plan and for the following step. How to get them: Getting a Request ID | Cursor Docs
Clarify if it reproduces:
on a different model in Plan mode
with MAX mode disabled
after restarting Cursor
Also share the approximate time of occurrence and the frequency.
Model/Mode: Reproduced with MAX mode disabled (standard Plan mode).
Restarted Cursor: Yes; still reproduced after restart.
Occurrence count: Now 3 times (twice with MAX on, once with MAX off after restart).
Request IDs:
Plan creation (MAX on): de4467a0-46fc-48cb-88e3-3accd9b0e165
Post-restart, MAX off: febf757e-c399-40ed-9caa-724c504aba33
Approx time of latest occurrence: (add your local time/UTC for the MAX-off run)
Console logs: Captured at the moment the state changed; available in logs/console-log.md (can provide a specific excerpt/timestamp window if you want).
Let me know if you want me to:
Try a different model in Plan mode,
Capture the exact Request ID for the MAX-off run,
Trim and attach the exact console snippet around the state change.
Reload the window and promptly click the “Build” button before the window displays “Built.” Refrain from utilizing the GPT-5.1 Codex Max model in “Plan” mode due to this bug.