Claude Code only works while it's open? Its built-in scheduled tasks fix that

Share

Claude Code has three kinds of scheduled tasks: cloud scheduled tasks, Desktop scheduled tasks, and /loop inside a session. Compare where each one runs, whether your computer must stay on, and the shortest interval. A prompt for your AI agent to set one up is included.

Machine translation. Read the original

SourceAnthropicReported Nov 15, 2026Not yet tested

Claude Code does nothing when it is closed, and it stops in the middle of a task and waits for you. Its built-in scheduled tasks can handle this, so you do not need another bot to nudge it. According to the official docs, there are three kinds of scheduling. Cloud scheduled tasks still run when your computer is off. Desktop scheduled tasks can use local files. /loop inside a session is good for watching one thing for a while.

Which kind to choose

Cloud scheduled tasks Desktop scheduled tasks /loop
Where it runs Anthropic's cloud Your computer Your computer
Does your computer need to stay on? No Yes Yes
Does the session need to stay open? No No Yes
Can it read local files? No. It clones the repo fresh each time. Yes Yes
Shortest interval 1 hour 1 hour in the form. Claude can set it down to 1 minute. 1 minute

The official advice is direct. If you need reliability and do not want to depend on your computer, use cloud scheduled tasks. If you need local files and tools, use Desktop scheduled tasks. For quick polling inside a session, use /loop.

  • Cloud scheduled tasks used to be called routines. They are still in research preview. A task is a saved setup. It includes a prompt, the repo to use, and a set of connectors. Besides running on a schedule, a task can also be started by an API call or a GitHub event. They are available on Pro, Max, Ultra, Team, and Enterprise plans. Manage them at claude.ai/code/routines, or use the /schedule command in the terminal.
  • Desktop scheduled tasks start a new session automatically at the set time. They suit daily code checks and morning briefings. They only run while the app is open and the computer is awake.
  • /loop is written as /loop 5m check the deploy. If you leave out the interval, Claude decides how often to check.

One user report describes using another chat bot to control Claude Code on the computer on a schedule, to keep it writing code. This approach works, but it adds one more layer. The other bot's own usage limits and stability then become the bottleneck. This is one person's experience and is not verified yet.

Ask your AI agent to set it up

Send this to your AI agent:

I want this project to keep moving forward a little every day. Set up a scheduled task for me:
1. First ask me three multiple-choice questions: Will my computer stay on all the time? Does the task need to read local files? How often should it run?
2. Based on my answers, pick one of cloud scheduled tasks, Desktop scheduled tasks, or /loop, and explain why.
3. Write the task prompt: each time, first read the TODO list and only do the top item. After finishing, run the tests and write the result in the progress log. For any delete, publish, or spending action, stop and leave a note for me.
4. Set up the task, then tell me where I can see when it runs next and how to pause it.

Notes

  • No one is watching a scheduled task, so decide on permissions first. Cloud tasks use Auto Mode by default to handle permission requests. For Desktop tasks, you can set permissions for each task.
  • Every run uses your usage. If the interval is too short, or the prompt is too broad, your usage can drain without you noticing.
  • Tasks made by /loop depend on the session and expire after seven days. For long-running work, use one of the other two kinds.
  • If you want Claude to keep working toward a goal until it is done, rather than repeating on an interval, the official docs describe /goal for that.

Official docs: Run prompts on a schedule and Automate work with routines.

Done with this one? Mark it to track your path.Sign in
Save it for laterSign in
Questions about this piece?Ask AI about this

Comments (0)

No comments yet. Start the conversation.

Sign in to join the discussion.Sign in