There’s no longer an unlimited agent‑request “Pro Mode.” Previously, as captured in my screenshots, once you hit the usage limit of the current model, you could wait a bit and then resume. But now, you must wait until the next billing cycle to use it again. This disrupts workflow continuity—it forces us to either pay more or switch to a different model mid‑task. As you know, such interruptions break the development flow and are something developers generally dislike.
I understand the unlimited agent requests were a huge draw and popular among many users, so it’s surprising that they were removed without any announcement. It would’ve been nice to receive some kind of cursor-like message—similar to Netflix notifying you when a movie is about to expire—perhaps via email.
I suspect the change may be related to new commercial agreements surrounding the AI model. But regardless, the sudden, unexpected switch caught me off guard.
Just to add, while you get an “uncapped” amount of requests each month, we found the wording of “unlimited” to be confusing for people on the new pricing system, so we reworded things to make it clearer how the system worked.
The system has not changing, and you can still use an unlimited amount of requests, but you may still be rate limited depending on the model you choose to use.
As @aldinokemal mentions, if you use ‘Auto’, this is currently truly unlimited, so it is free to use as much as you like. However, using a model like Claude 4 Opus is a very expensive model to run, so you would likely hit the rate limits very quickly.
I agree—Claude 4 Opus is pretty pricey, and Sonnet 4 isn’t cheap either. But it feels like walking into an all-you-can-eat buffet only to be told that because one particular dish costs too much to make, you suddenly have to take a smaller portion.
Cursor you guys should learn from anthropic , they had a billing issue not only they refunded but they provided 20usd api credit for every user as an apology
Previously it literally said “Unlimited agent requests.” I used it daily for hours and it worked fine after cooldown. Now it just cuts me off mid-task and says “access capped until next billing cycle”? That’s not “unlimited”—that’s 20-dollar capped.
Really disappointing how this was changed without a clear notification.
I dont understand rate limiting on model selection itself. We have hit the rate limit of sonnet 4 but why is it preventing us from using Grok 4. Which was recently launched on the platform.
