I’m on an individual Ultra plan. My First-Party Models usage is at 22% and my API usage is at 100%.
Bugbot now refuses to review PRs with “usage limit reached”. On-demand spending is disabled.
Cursor’s documentation say that individual Bugbot reviews first consume “included Cursor usage”. Can Bugbot consume either included pool, or is it restricted to the API usage pool?
If it is restricted to API usage, could Cursor please document this explicitly?
Steps to Reproduce
Try to use Bugbot when I have available first-party model usage on my plan but no remaining API usage.
Hey there!
Individual Bugbot reviews are restricted to your API usage pool. They draw from your included API usage, not your first-party model allocation, and the two don’t cross over for Bugbot. So even with first-party headroom left, once your included API usage is used up and on-demand spending is off, Bugbot stops with “usage limit reached.” That lines up exactly with what you’re seeing.
To get Bugbot running again, you have two options:
Enable on-demand (usage-based) spending in your dashboard. Bugbot will then continue past your included API usage and bill the overflow to on-demand.
Wait for your next cycle, when your included usage resets.
If cost is the concern, you can also dial Bugbot down so each review uses less: set it to run only once per PR or only when mentioned (bugbot run / cursor review), and lower the effort level. Those are all under Bugbot configuration.
On your second point, you’re right that this isn’t spelled out. The pricing docs just say Bugbot “first consumes your included usage” without clarifying that for individual plans that means the API pool specifically, not first-party. That’s a fair gap, and I’ll get it clarified.
It seems like a real product gap though that is inconvenient at best and, at worst, is using up paid API credits unnecessarily. Bugbot now also runs in the IDE and cloud agents where it runs on first-party models. However, Bugbot in the cloud through GitHub now can only be run via API usage (which has become a more precious commodity now with the expensive frontier models).
For my use case, I can just turn off Bugbot and rely on other forms of review before I commit. But that takes away a great feature that I loved with Cursor, which provided me peace of mind as an automatic line of defence in my PRs.
I’ve used Cursor almost daily for the past two years and subscribed to Bugbot as soon as it launched, which I was comfortable paying an additional (capped) fee for. I was also comfortable when it was included in the main subscription, where I could choose the model it used.
In that context, the current configuration feels like a significant step backwards and may push users away from Bugbot.
Could you please raise this with the product or pricing teams for consideration?
I was also caught out by this. Especially since it was announced that BugBot used Composer, which made me think it would consume Auto/First-Party credits
This is very annoying, same for the security automation as well. I had to turn it off as it used 7% of my Ultra account in a single evening. Not being able to adjust the model, not having it clearly mentioned and it not following all my default model configurations feels like a broken feature.
