One solution found**
**
Not sure if this helps but after Cursor has been unusable for nearly a week on my end, a lot of patient troubleshooting with a colleague we have resolved the issue.
Recap, experiencing:
No issues using VS Code
SSH was the issue
Agent chat “hung” on “planning next moves”
Diagnostics tests all green
Agent worked locally
HTTP1.1 did nothing
After many many many things we checked out the Indexing Tab and I had an error (not shown) but something akin to “cannot index”. We then reopened a smaller subsection of our repo and voila - agent chat started working again.
I’ve reproduced it - as long as Codebase Indexing is green - agent chat works.
If I open the project in WSL2, the agent works fine.
If I relaunch it in the dev container, it does no longer work.
I have confirmed the dev container has access to the internet without any issues and is up-to-date.
The dev container I use for this project has the same (official microsoft) base container as the one that is working fine.
Is there any way to investigate this issue from the CLI?
Update: The exact same project and dev container runs fine when I open it at home.
The only difference is that I am still using the old microsoft dev containers extension at home (as it is still far superior to the new anysphere one) and a different ISP.
I would love to pinpoint the exact cause of the issue, but need to know what the ‘network diagnostics’ check actually does so I can reproduce it in the terminal to find the cause.
Update 2: It now works again without any issues from both locations.
I have no explanation why it fixed itself, I have not changed anything ¯\_(ツ)_/¯
It’s made Cursor completely unusable for 4 of 7 members of our team.
We’re now on 4th-5th day of this problem. Most of us are using Claude Code as the workaround, but we may end up migrating permanently if this goes much longer…
Since today can’t use cursor anymore. If I enter a prompt for Agent it stucks on “planning next move” forever. If I stop it after 5min it still used up credits but did nothing. No Output. Even if the request is very simple.
One RequestID for example: feb2158f-f60a-4429-b526-99847c61cda6
Tried in “Auto”, “Haiku 4.5”, “Composer 1”
Even if I make it very simple “Create a lorem ipsum answer” as only prompot without any @ mention of a file it got stuck forever. fdd0b99c-993c-4fc5-902c-8f284bd9128e
I think it have to do something with the cursor files in the project. If I start cursor on a blank text file it is slow as ■■■■ but it works. (For example “create a html
element with lorem ipsum” on a blank file took it nearly a minute.
Steps to Reproduce
run any prompt on an existing project.
Operating System
Linux
Current Cursor Version (Menu → About Cursor → Copy)
I have the same problem. After the update, the cursor began to think about the answer for a very long time. It is now impossible to work in the ‘auto’ mode.
Same thing here. I had lots of repos open and therefore I suppose the indexing couldn’t finish which was freezing any process from cursor (tab, network…).
I just closed a bunch of repos and restarted cursor and it fixed itself.
Why did you guys remove the output logs from the cursor background agent? This issue is most likely caused by issues with install scripts but I can’t see anything after the latest update so I have no idea what docker build or the install script is logging..
It’s a bad plugin/extension (likely one that hasn’t been tested or enhanced to support the latest Cursor UI version). I disabled all my extensions and only added back the few I felt I needed. I left the rest uninstalled. Quit the app and restarted and now it works.
