Others are saying that 2.3.35 is the latest without the problem, but I’m on 2.3.41 and can attest that it too does not have this issue.
Has anyone on the Cursor team acknowledged that they are aware of this issue (and ideally that they’re working on it)? Cursor has been completely unusable on linux for a month unless the user knows to downgrade.
This issue is back for me after updating to version 2.5.17 - I’m not sure how to roll this back on a mac. It says no updates are available in beta or the regular updates.
i also **run into**this problem on windows11 with cursor version 2.5.26, but i deal with it by waiting about 5-10 mins then is works well as before ,i gusee may it is a progress that cursors‘ agent is initializing.so may u can try wait 5-10mins then try again to see if the problem can solve or not
This is insane how Cursor became a unusable product where you just have to get used to navigate countless of errors, and where updates make things worse instead of improving/fixing them.
It’s strange. I use Cursor on both Mac and Windows. The problem started after I updated to the latest version on Windows, and after that, even when I rolled back to a previous version, it still didn’t work. On Mac it continues to work normally, but I haven’t updated the version there. The Mac version is 2.4.37.
Hey everyone, sorry for the long silence in this thread.
This is a known issue with the extension host that started showing up in versions 2,4.x. The team knows about it and is tracking it. It affects all platforms.
I’ll let you know if there are any updates on this bug.
And they never finish. Also in Cursor settings in “Rules, Skill, Subagents” everything is stuck in a loading state.
I think the problem is that the rgcommand has */**/.cursor/rules/**/*.mdc as a pattern in conjunction with --follow which can quickly balloon the search space.
My hack for now is adding a .rgignore in my repo root that has something like:
*
!.cursor/
!AGENTS.md
This hack has fixed the problem for me. But it might have other side effects because rgis used by other tools/extensions as well. So hopefully there’s a official fix soon.
OS: macOS [e.g. Sonoma 14.x — replace with your version]
Cursor: [e.g. 2.6.19 — from Menu → About Cursor → Copy]
Behavior
The issue does not happen right after opening Cursor. It appears after leaving Cursor idle for a while (e.g. switching to another app or closing the laptop lid). When I come back, I see “Waiting for extension host” and it happens almost every time.
Disabling all extensions does not fix it, so it does not seem to be caused by a single extension.
From my local logs (for your triage)
In main.log: before/during the issue I see CodeWindow: detected unresponsive, followed by multiple Extension host with pid xxx exited with code: 0.
In exthost.log: Extension host terminating: renderer closed the MessagePort.
I also see No ptyHost heartbeat after 6 seconds and [backup tracker] suspended in the logs.
So it looks like after idle the window is marked unresponsive or suspended, the renderer closes the MessagePort, the extension host exits, and then the UI waits for a new one — hence “Waiting for extension host”.
Workaround
Cmd+Shift+P → “Developer: Reload Window” recovers the session until it happens again.