-
When cursor rules are applied, please put something in the chat window to indicate which rules were applied. It would greatly help in debugging and improving rules.
-
Likewise, when using “auto” select mode with an agent, it would be great to see which model was used at each point of the chat. This would help us give you guys much better feedback and also improve confidence in using auto mode.
-
Add a front-matter setting in cursor rules mdc to request the use of a specific model when the rule is applied.
2 Likes
+1 
Big +1 for this. I’ve been using auto mode for the past week for science. I learned the hard way that “auto” mode can and will use 3.7 Max, which eats up requests a LOT faster…
I was frustrated to see this, because the stuff I was working on has been pretty simple. Nothing that would have been pushing the limits of claude-3.5-sonnet.
Given that when Max was announced, there’s a pretty big caveat of it costing a lot more and being an option for the extreme users who need to push the AI to the limits. It feels a little scummy to me that auto mode can and will frequently select it, especially if it’s a smaller task.
Especially because I did not choose to enable the Max model in my settings, and I’ve never used it intentionally. I disabled it in my models for now to prevent this.
Hi Would you mind posting here the Request ID for those MAX requests? it is not supposed to call max if you are on auto and Cursor team should have a look.
Some use the quick-fix by adding in the mode advanced settings a instruction to state at begin of output the model name (or otherwise into a rule file)
I don’t know how to tell which requests MAX was used for 
Chat history and individual chat messages are very vague and don’t give any indication of timestamps or which model was used.
The usage-based billing log on Cursor Settings should really show the associated request ID.
1 Like
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.