Keeping Claude Code working when you walk away

You give Claude Code something that will take twenty minutes. You need to leave. Everything about the situation is ordinary, and the default outcome is that the work dies with your session.

Three separate things can kill it, and they are worth keeping apart because different fixes address different ones:

• The process stops — the machine sleeps, or the shell that owned it exits.
• The connection drops — you lost signal, or your phone suspended the app. The process may be fine; you just cannot see it.
• The session is forgotten — the process survived but nothing knows how to reattach you to it.

1. tmux on a machine that stays on

The classic answer. tmux owns the process, so detaching your terminal does not kill it:

tmux new -s work
# start claude, then Ctrl-B then D to detach
tmux attach -t work

This solves the connection problem and the forgetting problem outright. It does not solve the first one: tmux on your laptop still dies when your laptop sleeps. It only works if it runs on something that stays awake — which in practice means a VPS or a desktop you leave on. See running Claude Code on a VPS.

The catch on a phone is ergonomic rather than technical: you are reattaching to a full-screen terminal UI and driving it with a soft keyboard.

2. A session daemon on the server

Same idea, one layer up. Instead of a terminal multiplexer, a small long-lived process on the server owns each Claude Code session and keeps a transcript. Clients attach and detach; the session neither notices nor cares.

The practical gains over tmux are that reattaching is instant and stateful rather than "here is a screen buffer, find your place", several devices can watch the same session, and the daemon can push a notification when a task finishes — because something is still watching even when you are not.

This is how Maude works: a per-server daemon owns every session's process, so closing the app is a detach rather than a disconnection, and you get a push when the work is done.

3. Anthropic's cloud sessions

If you own no machine that can stay awake, move the work off your machines entirely. Claude Code on the web runs tasks on Anthropic-managed infrastructure, so a session continues after you put the phone away and with your laptop shut. Start it on a laptop, review it on a phone.

The constraints: your repository has to be on GitHub, the work happens in an environment Anthropic provisions, and it cannot see your private network or database. Compared in more detail in Maude vs Claude Code on the web.

4. Remote Control (with a caveat)

Remote Control connects the Claude app to a session running on your own computer, and it is included with a Pro or Max plan. It fixes the connection problem well — Anthropic notes that if the machine sleeps mid-session, the session reconnects when it comes back online.

It cannot fix the first problem, because it is definitionally driving your machine: that machine has to stay on with Claude Code running. If the laptop in question is the one going into a bag, this is not the answer for that laptop. It is an excellent answer if you have a desktop at home that is always on.

Choosing

One thing that helps regardless

Whatever you choose, a task you cannot supervise should be written so it does not need supervising. Say where to stop, what to do on failure, and what not to touch. "Fix the failing tests in api/; if a fix needs a schema change, stop and explain instead" survives being left alone far better than "fix the tests" — the specific instruction is what stops a twenty-minute unattended run from wandering somewhere you did not want it.

Frequently asked questions