It doesn’t matter how many ways you say Proceed or Yes it always seems to ask some sort of variant question
Is there anyway to let Cursor know, stop asking me the question, just go ahead with the fix?
I have looked in the settings and i cant see anything to turn this off.
I’m having the same issue. I even used the new /create cursor rule and told it to create a rule to just proceed and never ask again but it kept asking for permission even for small insignificant changes. here’s the rule:
The AI must never ask the user to proceed or confirm before taking action. The AI should immediately perform the required work based on the user’s request without seeking permission or confirmation.
Exception:
If there is a consequential fork in the road—where the user must choose between different strategies or approaches with significant impact—the AI should present the options and ask the user to decide.
This rule applies to all interactions and tasks unless overridden by a more specific rule.
Very good solution, would the rule be alwaysApply: true so its always included?
Yeah actually it should be true. The problem is that I set it to true and it’s STILL asking for permission. My current frustrating solution is to type “PROCEED WITHOUT ASKING PERMISSION” at the end of each chat msg 
it could be a bug or by design. It’s profitable for Cursor to make you make 2 requests for a single task.
I had this issue and it drove me insane lol. Make sure you don’t have auto-select on for models or your requests will get sent to gemini, which does this.
There was some update that added auto-select as the default setting.
If you’ve been a longtime cursor user and thought you had your favorite model selected, it is a big surprise.
Yeah that was some versions ago due to overloaded AI API providers. But now if you set a model it sticks.
On which model/models does this happen for you mostly?
I found some models like Cl3.7 and Ge2.5pro fond of being too eager or too literal.
If you dont tell it to do something it will ask trying to be helpful (e.g. How can this feature be implemeted? → Would you like me to…)
If you tell it to do something, it may overdo it 
I was using gemini 2.5 for a long time and its fast and proceeds without asking me, life was good. But I recently switched to the default sonnet 3.5 and now it’s a hit or miss, sometimes it proceeds immediately and most times writes an essay with a “would you like me to …” conclusion at the end. So im not sure if there’s some shared memory cache or something went wrong and im inclined to remove cursor entierly and reinstall it.
I’m gonna switch back to gemini 2.5 and see how it behaves.
Hmm im using 3.5 mostly too because its following instructions better.
Curious if that is because of prompts/rules being needed to be slightly different per model.
Only when it starts running out of context length it starts forgetting parts and asks then for permissions, but thats on a long chat, with many actions.
I’ve been using gemini 2.5 for the past couple hours and it proceeds to make changes immediately, wonderful! I suppose Claude 3.5 sonnet was the culprit in my case.