I have been experiencing significant issues with Cursor AI for about a week now. I primarily use Composer 2.0 to work on my project, but it has become almost unusable.
Even when I ask it to perform a very small task, I constantly see messages like “Taking longer than expected etc.” The performance is extremely unstable and slow, making it impossible to maintain a professional workflow.
I am a Pro+ subscriber, yet the program is currently not fit for purpose, and I am being forced to use competitors’ tools to get my work done.
I have already checked my network diagnostics (which are successful) and tried re-indexing my codebase, but the issue persists. Please help me resolve this as soon as possible.
Steps to Reproduce
Open a project with around 1500 files.
Open Composer 2.0 (Cmd+I / Ctrl+I).
Select any model (e.g., Claude 3.5 Sonnet or GPT-4o).
Provide a simple prompt for a small change in the code.
Observed result: The generation hangs, and a “Taking longer than expected” message appears. The process is extremely slow or fails to complete, despite having a Pro+ subscription and a healthy network connection.
This happens consistently across different small tasks within the project.
Hey, thanks for the report. This is a known issue, and we’re seeing similar reports from other users.
For debugging, we need the Request ID from one of the stuck requests. Here’s how to get it:
Create a new chat and send any prompt via Composer 2
When the response finally comes back or gets stuck, click the chat menu in the top right, then select Copy Request ID
Paste it here
Also try this workaround: open settings with Cmd+, search for HTTP/2, and turn on Disable HTTP/2. Then restart Cursor. In some cases, this fixes delay issues.
The team is aware of this issue. The Request ID will help us narrow down what’s happening in your specific case.
I am also having the same issues. Tasks that used to be lighting quick take 10 minutes+ now. And ive switched back to Composer 1.5 and same issue. For the cost of the Pro+ subscription, this is ridiculous. It complete ruins the flow of development. And since AI often hallucinates the wrong solution, fast iteration is imperative. But due the extreme slowness over the last few weeks, iterating is painful and often i wonder if I could just write the code faster my self. And Im not trying to be dramatic, truly, i could do a lot of this with higher accuracy and speed when you account for how many iterations it takes to complete a task.
I‘ve switched model from composer 2 to opus. Works much better. Composer worked very well after releasing, efficiency of that model is not acceptable now..
The chat hanging / ‘Taking longer than expected’ issue has been addressed in a recent Cursor update. Updating to the latest version should resolve this.