General
Share and discuss tech, products, business, startups, or product recommendations
Start new thread

Is it actually a problem that your AI agent stops running when you close your laptop?

by
•

I’ve noticed that many coding-agent workflows still depend on keeping a laptop awake and connected.

I’m curious how people here handle longer tasks:

Do you leave your computer running, use a remote machine, or rely on a cloud environment?

I’m also interested in collaboration. When several people are working on the same project with different agents, how do you avoid conflicts and keep visibility into what each agent is doing?

Is this a recurring problem in your workflow, or something you rarely think about?

60 views

Add a comment

Replies

Best

Sounds like real problem , hunn

I think you should run agents on a cloud VM or VPS with tmux, or you can use Github Actions runners.

After that you should be able to close your laptop freely.

The issue is not agent stopping, it is context loss when you come back. Try slack summarise and commit checkpoints.

I think that can help you

 You are right about the real cost, and it took us a while to see it. The agent stopping is annoying, coming back to a 4000 line scrollback and not knowing what happened is what actually wastes the afternoon.

tmux on a VPS is exactly where we started and it holds fine for one person. Two things broke it for us. A second person attaching to the same tmux session shares

one cursor, so you are fighting over the keyboard instead of working. And several agents in one checkout overwrite each other silently, which no amount of persistence fixes.

  

On context loss, commit checkpoints are half of it and we do the same, plus a push before a machine is ever torn down. The other half that helped more than expected: each agent writes a short summary at the end of its turn, and when you come back you read a diff of what happened since you last looked rather than the transcript.

Same thing the other agents get, which is also how they stop redoing each other's work. 

  

That is what we ended up building into Murmell, though the tmux route is genuinely the right first answer for one person on one machine.

i started a remote machine for longer tasks. it gives me pecae of mind when i need to step away.

hmm how do you handle security of this ?