In cursor IDE, agents stops working while Cursor Desktop app closes eventhough the agents run in remote machines .
I use Cursor IDE in Mac. All my work is in remote machine connected via SSH. Ongoing work in agent chats gets interrupted/coses when I close the app. Ideally the work should continue because these should be running in remote machine.
very frustrating to work on IDE. I only work with CLI now because of this. Codex and other harness handle this seemlessly. given a choice, users switch the harness for ease of use.
Steps to Reproduce
open cursor installed in Mac.
connect to remote machine.
open folder to any project workspace. start any agentic chat to do some work
close the app. open the app again later to continue the work- but doesn’t work the agent get stopped.
Hey @Ranjith_Poliyedath, this isn’t something wrong with your setup. Today a chat started in the editor is run by the desktop app, even on a Remote SSH workspace. The commands and edits happen on the remote machine, but the app drives the session, so quitting it stops the turn.
For now, the CLI in tmux on the remote machine is the reliable way to keep a long task going. In the desktop app, the Agents Window can pick an interrupted agent back up if you reopen within a couple of hours (Cursor Settings → General → Continue Interrupted Agents, on by default). That resumes the work when you come back, it doesn’t keep it running while the app is closed.
Cloud Agents are another good way to keep agents alive even after your laptop is closed.