Cursor IDE resource usage super high and crashes. [Linux]

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

When I try to use Cursor IDE , I can see my CPU, RAM spike super high even if I got pretty decent hardware, and then Cursor crashes with no error message … It’s doing this since like last week or something like that … I cannot use Cursor at all because of that …

Steps to Reproduce

Start Cursor, wait , send a message, anything…

Expected Behavior

Cursor IDE to work fine.

Operating System

Linux

Version Information

Version: 3.25.9
VS Code Extension API: 1.128.0
Commit: 5d4df965746c5d09740842fc0f1331ba1c157830
Date: 2026-10-15T21:52:24.891Z
Layout: IDE
Build Type: Stable
Release Track: Default
Electron: 40.10.6
Chromium: 144.0.7559.236
Node.js: 24.15.0
V8: 14.4.258.32-electron.0
xterm.js: 6.1.0-beta.291
OS: Linux x64 7.2.11-1-cachyos

Additional Information

I got a screenshot with CPU spiking at 99% and ram at 80% and I have a AMD Ryzen 7 5800XT and 40GB RAM DDR4 3200mt …

Does this stop you from using Cursor

Yes - Cursor is unusable

Hey @PerkyZZ,

Thanks for the version details. CPU and memory climbing until Cursor closes with no error message usually means one Cursor process is growing out of control and Linux then kills the app. We need to find out which process it is.

A few things to try, in order:

  1. Update to the latest 3.26 release (Help > Check for Updates, or download it from Cursor · Download). It includes changes that reduce how much agent conversation state stays in memory during long sessions. We haven’t confirmed it fixes your case, so please tell us whether it changes anything.
  2. Make sure only one Cursor install is running. If you have both a package-manager build and a downloaded build on CachyOS, quit both, check that no cursor processes are left (ps aux | grep -i cursor), then start just the updated one.
  3. Temporarily turn off MCP servers you don’t need (Cursor Settings > Tools & MCP). Each enabled server starts its own process in every window, so a long list adds up quickly.

I did update to 3.27 , also turned all my MCP off and the issue is still there … Here is a screenshot of the process explorer opened with my resources usage at the top middle …

Also a screenshot of my System Monitor …

Exact version I am on now is :
Version: 3.27.6
VS Code Extension API: 1.128.0
Commit: f7119520544accb8dcf6d27119b9f1963d8ee5a0
Date: 2026-10-21T09:08:09.275Z
Layout: Agent Window
Build Type: Stable
Release Track: Default
Electron: 42.11.18
Chromium: 148.0.7778.280
Node.js: 24.19.0
V8: 14.8.178.38-electron.0
xterm.js: 6.1.0-beta.291
OS: Linux x64 7.3.0-3-cachyos

Also Is there no way of clearing the cache and all my data from my Cursor ? I’d try that if possible to see if the issue comes from having too much threads , projects, etc… or if it really comes from the app itself …

Hey @PerkyZZ,

Thanks for the screenshots, they help a lot. They show the CPU isn’t being used by Cursor’s own processes. Everything Process Explorer lists adds up to about 4% CPU and roughly 5 GB of memory, while your system is at 99%.

Process Explorer only shows processes that belong to the Cursor window that’s open right now, so the load is coming from something outside that list. Most likely it’s processes left over from an earlier Cursor session (for example MCP servers or tools that didn’t exit when Cursor closed or crashed), but it could be something else on the system.

Also, MCP may not be fully off yet. Servers that come with installed plugins (such as shadcn and Serena) are separate from the ones in your own MCP list, so please turn those off too to rule them out.

Next time CPU climbs, could you run these in a regular terminal and paste the output?

  1. ps -eo pid,ppid,etime,pcpu,pmem,args --sort=-pcpu | head -20
  2. For the top three or four processes, run ps -o pid,ppid,args -p PPID, using each one’s PPID from step 1, to see what started it.
  3. Quit Cursor completely (Ctrl+Q), wait 10 seconds, then run pgrep -af -i 'cursor|bun|serena|uvx|npm exec'. Does anything still show up, and does CPU drop once Cursor is closed?

If Cursor closes on its own again, these would show whether Linux killed it:

  1. journalctl -k -b | grep -i -E "out of memory|killed process"
  2. journalctl --user -b | grep -i oom
  3. coredumpctl list | grep -i cursor

Also, before reopening after a crash, zip the logs folder (Command Palette > Developer: Open Logs Folder) and attach it here.

In the meantime, if step 3 shows leftover processes, kill PID (or kill -9 PID if it ignores that) clears them, and logging out or rebooting clears them all.

On resetting Cursor: yes, you can test with a clean profile. With Cursor fully quit, rename the data folders instead of deleting them, so you can put them back:

mv ~/.config/Cursor ~/.config/Cursor.backup
mv ~/.cursor ~/.cursor.backup

Then start Cursor and sign in again. It will start with default settings, no extensions, and no local chat history. If the problem goes away, the cause is in your old profile. If it doesn’t, rename the backups back to restore everything. I’d do the process check above first, though: since Cursor’s own processes look quiet in your screenshot, a reset may not change much.

Hmm just updated to the latest version and yesterday before going to sleep I did the Developper: Clear old discussion action in the Command palette and deleted all that was 14 days or older … When I start the app now the usage is fine, but when I send a message to my agent it gets to 90% + CPU usage … I haven’t tried moving my profile and try a clean one, that is what I am gonna do next and I will look at the logs also when I got time …

Hmmm I got the app to crash with OOM and the only log I could get was this one …

Oct 23 16:15:36 CachyOS-x64 systemd[1057]: app-cursor-3619562.scope: systemd-oomd killed 382 process(es) in this unit.
Oct 23 16:15:40 CachyOS-x64 systemd[1057]: app-cursor-3619562.scope: Failed with result 'oom-kill'.

Hey @PerkyZZ,

Thanks for the oomd log. One correction to my last reply: your System Monitor screenshot shows Cursor itself at about 73% CPU, so the load is coming from Cursor or the commands its agent runs, not from another app.

Two things would help most:

  1. If you installed from the AUR or a CachyOS package, try the official build from Cursor · Download (the AppImage runs alongside it). Packages that use the system Electron have caused crashes on Arch-based systems.
  2. If it still happens, send the agent a message and, while memory climbs, paste the output of:
    ps -eo pid,ppid,rss,pcpu,args --sort=-rss | head -20
    ps -eo comm= | sort | uniq -c | sort -rn | head

Alright, the only way I got the app to work properly is to move my profile and start from a new profile … That way it worked and it’s not using all my resources anymore … Would’ve been great to be able to keep my stuff but it’s fine like that it’s cleaner hahaha