I’m a Pro user on Cursor, and over the past few weeks, the platform has become increasingly frustrating to use.
First, there was a silent reduction in the premium model quota — from 500 to just 200 monthly calls to Claude 4 Opus, which is the only model that still performs decently. This already made things difficult, especially with no clear communication from the team.
The only alternative was to rely on Auto Mode, which used to work fairly well until July 22, 2025. But after that date, its performance has dropped significantly:
- The agent no longer executes simple tasks as it did before.
- It often responds with generic questions or comments instead of taking action.
- In many cases, it fails to apply file edits properly, returning
"cancelled"(Linter error) and becoming severely limited in practice.
To make matters worse, once the 200 premium requests are used up, no fallback models are unlocked, leaving Pro users without any functional options — which completely defeats the purpose of paying for the plan.
It’s frustrating to see a tool with so much potential becoming more and more limited, with little transparency or guidance from the Cursor team. I hope this is addressed soon, as the current situation makes Cursor nearly unusable for professional workflows.