Claude Code Can Now Run Scheduled Tasks in the Cloud Without Your Machine

The biggest complaint about Claude Code scheduled tasks has always been the same: your computer has to stay on. Desktop scheduled tasks are persistent and survive app restarts, but if your MacBook lid is closed or the app isn’t running, the task gets skipped.

That changes today. Anthropic just shipped cloud-based scheduled tasks for Claude Code.

Open post on X

Noah Zweben from Anthropic announced the feature, which lets you configure a repo (or multiple repos), a schedule, and a prompt. Claude runs it on Anthropic’s cloud infrastructure at the cadence you set. Your local machine can be off, asleep, or on the other side of the planet.

What Problem This Solves

Claude Code has had scheduling capabilities since early March 2026, but every option came with tradeoffs.

The CLI’s /loop command creates session-scoped cron jobs that die when you close your terminal. They also auto-expire after three days. Useful for polling a deployment during an active coding session, but not much else.

Desktop scheduled tasks are more durable. They persist across app restarts and support hourly, daily, weekly, and weekday cadences. But they only fire while the Claude Desktop app is open and your computer is awake. The docs say it plainly: if your machine is asleep at the scheduled time, the run is skipped. When your machine wakes up, Desktop runs one catch-up job for the most recently missed window, but anything older than that is discarded.

For developers who wanted true always-on scheduling, the official recommendation was to set up GitHub Actions with a schedule trigger. That works, but it requires maintaining a separate workflow file, managing secrets, and context-switching out of the Claude Code ecosystem entirely.

Cloud scheduled tasks fill the gap between Desktop’s convenience and GitHub Actions’ reliability. You get Anthropic-hosted execution with the same setup experience as Desktop’s local scheduling.

How It Works

The setup follows the same pattern as remote sessions in Claude Code Desktop. When starting a session, you choose “Remote” instead of “Local” as your environment. Remote sessions already run on Anthropic’s cloud infrastructure and continue even if you close the app.

Cloud scheduled tasks extend this by adding recurring execution. You provide three things:

A repo (or repos): Point Claude at one or more GitHub repositories. Multi-repo support means you can run tasks that span codebases, like updating a shared library and its consumers in the same run.
A schedule: Choose how often the task fires. The same options available for local scheduled tasks apply here: hourly, daily, weekdays, weekly, or manual.
A prompt: Write the instructions Claude follows on each run. This is the same prompt format you’d use in any Claude Code session.

Each scheduled run spins up a fresh cloud session. Claude clones the repo, executes the prompt, and can make commits, open PRs, run tests, or anything else it would do in an interactive session. You check back whenever you want to review results.

What You Can Do With This

The practical use cases break down into a few categories.

Overnight code maintenance. Schedule dependency updates, linting passes, or test suite runs for off-hours. Claude opens a PR with the changes and you review it over coffee.

Daily briefings. A prompt that reviews overnight commits, summarizes open PRs, and checks CI status across your repos. Claude can use connectors to post the summary to Slack or create a Linear issue.

PR babysitting. Set Claude to check on open pull requests every few hours, review new commits, flag failing CI, and post review comments. This existed with /loop before, but it died when you closed your terminal. Cloud scheduling keeps it running indefinitely.

Multi-repo coordination. Tasks that touch multiple codebases can now run unattended. Update an API contract in one repo and adjust the client code in another, all in a single scheduled run.

How This Fits Into the Bigger Picture

This announcement lands one day after Anthropic shipped Claude Code Channels, which lets you control Claude Code sessions remotely from Telegram and Discord. The pattern is clear: Anthropic is systematically removing the requirement that you be sitting at your computer to use Claude Code.

Two weeks ago, the /loop command and cron tools landed in v2.1.72. Last week, Desktop scheduled tasks got catch-up behavior for missed runs. Now cloud scheduling removes the machine dependency entirely.

The progression makes sense. Local scheduling proved the demand. Desktop scheduling proved the workflow. Cloud scheduling removes the last friction point: keeping hardware running.

What You Need

Cloud scheduled tasks require a paid Claude subscription (Pro, Max, Team, or Enterprise). Cloud sessions share rate limits with all other Claude Code usage, so heavy scheduling on a Pro plan will eat into your interactive session quota.

Remote sessions run in isolated sandbox environments with network and filesystem restrictions. Git interactions go through a secure proxy that limits access to authorized repositories. For teams, admins control whether cloud sessions are enabled through the admin console.

What to Watch

A few things worth keeping an eye on as this feature matures:

Rate limit pressure. Every scheduled cloud task consumes the same tokens as a regular session. Teams running dozens of scheduled tasks across multiple repos could hit their limits quickly, especially on Pro plans. Max subscribers with higher rate limits are better positioned for heavy scheduling.

Cost transparency. Anthropic says there are no separate compute charges for remote sessions, which means the cost is bundled into your subscription. That’s simple, but it also means there’s no granular visibility into how much each scheduled task costs in token consumption.

Notification and observability. Local scheduled tasks pop a desktop notification when they run. How cloud tasks surface their results (email, in-app, Channels integration) will determine how useful they are for set-and-forget workflows.

This is the kind of feature that sounds incremental but changes how you use the tool. Scheduling a daily code review and forgetting about it is a different relationship with an AI coding agent than sitting down and prompting one every morning. Cloud execution means that relationship doesn’t depend on your hardware being available.

Sources: