Cursor updated mid-agent and ate my docker-compose diff
Left Claude Code refactoring a compose file on Mac while I grabbed coffee. Came back to the app restarting itself — auto-update — and the dirty buffer was just... gone.
Recovered most of it from ~/Library/Application Support/Cursor/User/History, but it dropped the redis healthcheck and a volume mount. Spent 40 minutes wondering why staging had no cache again.
Do y'all pause agents before Cursor updates, or is there a setting I'm missing?
5 comments
Join the discussion
Log in to comment.
same energy as when Make.com "upgraded" a scenario overnight and silently rewired my Slack filter.
i just close the agent window the second I see the update toast. not elegant. works.
Closing the window mid-agent also cuts the tool call mid-flight. We lost a half-applied terraform plan that way once. I pause the agent from the chat first, then deal with the update.
I put a hard kill on the agent before any update prompt now. Learned this the expensive way — lost a FastAPI migration once, only had the last successful tool call in the transcript.
History folder saved me twice. Still ugly.
hard kill is the only move that works for me. i also
git stashbefore i walk away from a long agent run now. lost a healthcheck once and staging redis died for an hour before anyone noticed.Had this on Windows last week. Cursor restarted mid-edit on a compose override and the unsaved buffer just vanished.
I turned off auto-update under Settings → Application. Still get the toast, but at least it waits for me. The History folder trick is real though — I keep that path in a sticky note now.