Hello,
I’ve been working on a project with YOLO mode enabled, and I’ve encountered an issue where the agent stops after 25 tool calls, requiring me to start a new composer. This happens consistently, often interrupting my work mid-task. It’s a recurring problem I faced multiple times yesterday.
I noticed the default behavior is to stop the agent after 25 calls since it’s considered an experimental feature. However, this limit is quite restrictive for complex workflows, especially when working on tasks that require continuous assistance.
I understand the suggestion to contact hi@cursor.com for feedback, but I’m curious how others are addressing this limitation. Are there any workarounds or settings adjustments to maintain a smoother workflow? Alternatively, is there a way to extend the limit for tool calls to prevent frequent interruptions?
Any guidance or recommendations would be greatly appreciated.
Thank you!
1 Like
I did reach out and got a pretty quick response from the team which was nice-- I’m just asking it to continue now, and re-adding stuff to context manually as it seems (anecdotally) to lose track of the original extra context provided. This is still a huge improvement over approving every command execution before YOLO mode 
It seems there are others with the same issue. I say continue and it just sits there Generating… cancel reenter prompt does nothing. seems there are others that have downgraded and it is working for them moving from 0.44.11 > 0.44.10 I think this might be what I have todo.
From what I can tell, Cursor is putting a hard limit on the YOLO. I can go back to a single prompt in the same chat and suddenly… I can resubmit without reverting and it will respond.
Am I the only one seeing this?
Unfortunately, the generating bug is one we’re seeing for a totally different reason and is not related to the feature you use or the use of YOLO mode.
We are actively working on resolving this at the moment!
