Claude Code can now run tasks while you sleep - here is what that means in practice

Anthropic has added background scheduling to Claude Code, and it is worth understanding what this actually means in practice.

You can now give Claude Code a prompt and a schedule: every morning at 6am, every hour, every Friday. It will run that prompt automatically without you being there. You describe the task once and it repeats. No manual triggering, no opening a session and typing the same thing again.

The practical examples are things like a morning briefing that checks your calendar and outstanding tasks before you start work, hourly monitoring of automation logs, or a weekly summary of what has been completed across your projects. Anything you currently trigger manually on a regular basis is a candidate.

There are two versions of this feature worth knowing about. The desktop app version runs persistently and survives restarts. If your computer was off when a task was due, it catches up when you reopen the app. The /loop command in the terminal is session-only and stops when you close it. For anything you want running reliably, use the desktop app.

One thing that does not get mentioned enough: these tasks run without your supervision, so you need to be deliberate about permissions. Tell Claude Code explicitly what it is not allowed to do, deleting files being the obvious one. You can set this in your project settings and Claude Code will help you configure it if you ask.

The other genuinely useful pattern is giving each scheduled task a memory file. Structure your prompt so that at the end of each run, it writes a brief status note to a file: what it did, any issues it found. The next run reads that file first. This means tasks can improve over time and stay aware of what happened previously, even though each session starts fresh.

This moves Claude Code from something you actively use to something that is also working in the background.

Another video for

@Ryan Doser

list😅

3 comments