Often in Plan Mode, and especially after answering questions, it will just hang and show “Taking longer than expected” even though it built the plan, though it often puts the plan in a location that is not accessible and when I ask for it to link me to the plan the link is broken.
Steps to Reproduce
Enter plan mode, tell it to ask you questions, then answer the questions.
The secondary issue of the plan not being created in an accessible place is definitely intermittent, sometimes it does put it in the right place, but when this “Taking longer than expected” issue happens I have to interrupt it and ask it to give me the location of the plan
This is a known issue. Plan Mode can sometimes hang with “Taking longer than expected” – particularly after answering questions – and intermittently place the plan output in the wrong location. Our team is actively tracking this, and a fix for one of the underlying causes (plans not being created properly) was just shipped.
A few things to try:
Update Cursor to the latest version (Help > Check for Updates). A recent fix addresses cases where the plan output ends up in the wrong place, which may help with the broken links you’re seeing.
Copy and re-prompt: If the plan gets stuck, interrupt it, copy the plan content from the chat, open a new chat (Ctrl+N), and paste the plan directly without enabling Plan Mode. The agent should then execute from there.
Try a different model: Some models handle the Plan Mode flow more reliably. Try Claude 4.6 Sonnet or GPT-5.5 for planning tasks to see if the hang frequency decreases.
If the issue continues after updating, it would help to have a Request ID from one of the stuck sessions (three dots menu at the top right of the chat > Copy Request ID). That would let us trace the exact failure point.
Using a different model is not an option. The only model that produces plans that meet our standards for consistency and quality is Opus 4.6. And we regularly re-test this with new model releases.
Understood – thanks for confirming. The update addressed a different variant of this issue, so it makes sense that the core hang is still happening for you.
To get this prioritized for a fix, the most useful thing right now is a Request ID from the next time it hangs. When you see “Taking longer than expected,” click the three dots at the top right of the chat and select Copy Request ID, then paste it here. That lets our engineering team trace exactly where the request stalls.
