Every prompt I hit in cursor ide, it stucks on “Taking longer then expected.” and never complete. I have been trying since Friday, and till now same. i tried relogin, force update the ide, but still same.
Steps to Reproduce
Simple, open IDE, open the chat window, add prompt and hit enter or send
I’m experiencing the same issue, and it’s taking much longer than expected to resolve.
I have installed the latest version on Windows 10, but Cursor is currently unusable. Every time I try to use the chat, it displays the message “Taking longer than expected” and then remains stuck indefinitely without generating any response.
The issue occurs consistently with every chat request, making it impossible to use the application.
Could you please advise on any workaround or provide an update on the status of this issue?
If Cursor keeps hanging on startup with the “Taking longer than expected…” message and you’re on Windows 10, the culprit might be Windows Defender — even if you’ve already added folder exclusions.
What worked for me:
Adding Cursor.exe as a process exclusion in Windows Defender, not just a folder exclusion.
Steps:
Open Windows Security > Virus & threat protection > Manage settings.
Scroll down to Exclusions and click Add or remove exclusions.
Folder exclusions prevent Defender from scanning files on disk, but they don’t stop Defender from inspecting the process’s network activity in real time. Adding the process exclusion is what actually prevents the hang on startup.
Optional but recommended — also add these folder exclusions for better performance:
Quickest unblock: fully close and reopen Cursor (not just reload the window) - a fresh window clears it when it gets stuck like this.
On Windows, the fix that’s held for others is @actuaan’s above: add Cursor as a Windows Defender process exclusion (not just a folder). Your Network Diagnostics shows streaming getting buffered, so if it’s still stuck after that, set Cursor Settings → Network → HTTP Compatibility Mode → HTTP/1.1 and restart.
This is an issue we’re tracking on Windows - I’ll post here when there’s an update. If it keeps happening, run Command Palette → “Developer: Capture and Send Debugging Data” right after it hangs and reply with the time and your timezone.
Same issue here - Windows 11 - service has been spotty for about 16 hours or so - bots are just hanging randomly and then 20 minutes later they start working again. Looks like backend to me - nothing to do with my machine.
I ran scripts to get http/2 working from the getgo, so I don’t think that’s related. Network Diag checks out fine. Claude specifically I’m only running in one window (I have a few agents going in parralel, all the rest are grok/composer). Keeps randonmly hanging - tried checking FAST mode to compensate - burned through tokens like crazy and still hung - datastreams are corrupted. When writing an image file to google drive which should have taken it no more than 2 minutes, it spent at least 20 minutes retrying over and over again and in the end I still ended up with a completely garbled image. Chewing up my other model usage because of backend issues… very frustrating.
Team is facing a lot of slowness while prompting and at the end gets timeout issue. Impacting the productivity and deliverables to meet the project timelines
Thanks all. We’re seeing “Taking longer than expected” on both Windows and Mac and are tracking it.
What clears it fastest: fully quit and reopen Cursor (not just reload), or open a fresh Agents window, since once a window latches into this state every message in it hangs. On Windows, adding Cursor as a Windows Defender process exclusion has held for others (@actuaan’s steps above). If Network Diagnostics shows streaming being buffered, set Settings → Network → HTTP Compatibility Mode → HTTP/1.1 and restart.
A few of you describe something that looks different: intermittent, only on one model (e.g. Claude while grok and composer work), or garbled output. That points at the request rather than the window. If that’s you, right after it hangs run Command Palette → “Developer: Capture and Send Debugging Data” or grab the request ID, and reply with it plus the time and your timezone. That lets us separate a stuck window from a request-level issue.
This is happening second time in span of a week.
First time I’ve cleared my user history and it helped because reinstalling an app and choosing different versions didn’t solve the issue. I don’t want to lose my data now.
I’ve cleared all of possible networking issues like firewall or vpns on my side.
I’m also encountering this issue inside of WSL. I ran through the steps mentioned previously by Alberto and it fixed it when I’m outside of WSL but not inside of it.
