after doing the research, plan mode does not create a plan but rather lists a message. this has happened twice already. see the screenshot to see the faulty header of the invocation method being written as part of the chat.
Hey, thanks for the report. It looks like a rendering/formatting issue where Plan Mode outputs raw invocation code instead of building the plan structure.
To help us investigate:
Request ID:
Please share the Request IDs from the sessions where this happened. Make sure Privacy Mode is disabled when you collect them. Getting a Request ID | Cursor Docs
Full screenshot:
Please send a wider screenshot showing the full chat context before and after the error.
Pattern:
Does this happen with certain types of prompts, or randomly?
Have you tried switching to a different model (e.g., gpt-5) to see if it reproduces there?
I am facing the same issue for a couple of days. Thanks for acknowledging the report.
Here’s the diagnostic information you requested:
Request ID c6dafd09-4252-47ac-a19e-bafb52f6190e
Full Screenshots
[Attached] Two screenshots showing:
UI status messages (“Thought for 21s” → plan preview → “Created plan”)
Editor maximized view with empty offline-plan.md file and Build button non-responsive.
Pattern
Occurs consistently across all prompts and all models (tested GPT-5-high-fast, others). No intermittent behavior—100% reproducible given the same project context and Plan mode request.
Additional Context
The UI displays the plan summary text and confirms “Created plan” status, suggesting content generation reaches completion. However, the draft persists as empty and Build does not execute, indicating a failure during the finalization/persistence layer after generation completes.
Environment: Cursor IDE (window), Nekobox VPN, no local changes. Other modes (Agent, Talk, etc.) function normally.
I’m available to provide browser console logs, network traces, or any additional telemetry the team requires. Let me know the preferred method for collection on the web client.
