Claude Just Got Scheduled Tasks: Your AI Can Finally Work While You Sleep

This is the feature I've been waiting for. Claude Code now supports cloud scheduled tasks—automated prompts that run on recurring schedules using Anthropic's infrastructure, not yours.

Translation: your AI can work while your laptop is closed, while you're asleep, while you're on vacation. It just runs.

Why This Matters

Until now, running AI tasks on a schedule meant keeping your machine on. Desktop apps, terminal windows, scripts running in the background—all dependent on your hardware staying awake.

Now? Claude handles it in the cloud. You set it up once, and it runs forever. No machine required.

How It Works

Creating a scheduled task is straightforward:

• Web: Go to claude.ai/code/scheduled and click "New scheduled task"
• Desktop: Open Schedule page → New task → New remote task
• CLI: Run /schedule in any session for a conversational setup

The CLI approach is surprisingly natural. Just type something like:

/schedule daily PR review at 9am

And Claude walks you through the rest—repos, environment, connectors, timing.

The Setup Steps

1. Name it and write the prompt. This is critical. The task runs autonomously, so your prompt needs to be self-contained. Tell Claude exactly what to do and what "done" looks like.
2. Select repositories. Each repo is cloned fresh at the start of every run. Claude creates claude/-prefixed branches for changes.
3. Choose an environment. This controls network access, environment variables (API keys, tokens), and setup scripts (like installing dependencies).
4. Pick a schedule. Hourly, daily, weekdays, weekly. Custom intervals available via /schedule update in CLI.
5. Review connectors. Your MCP connectors are included by default—remove any the task doesn't need.

Click create, and you're done. The task runs automatically at the scheduled time. Each run creates a session where you can review what Claude did.

5 Use Cases I'm Already Using

1. Morning PR Review

Every morning at 8am, Claude reviews all open PRs in my repos, summarizes changes, flags potential issues, and posts a summary to Slack. I wake up to a clean overview instead of an inbox of notifications.

2. Overnight CI Failure Analysis

CI fails overnight? Claude analyzes the logs, identifies root causes, and creates issues with reproduction steps. I start my day with actionable bugs instead of mystery failures.

3. Weekly Dependency Audit

Every Monday, Claude checks for outdated dependencies, security vulnerabilities, and breaking changes in my projects. Creates PRs with updates where safe.

4. Documentation Sync

After PRs merge, Claude updates the relevant documentation sections. Keeps docs in sync without manual intervention.

5. Content Pipeline

My content research agent runs daily, scans trending topics, and saves content ideas to my notes. When I sit down to write, the research is already done.

The Comparison Table

Claude now offers three ways to schedule work:

• Cloud tasks: Runs on Anthropic infrastructure. No machine needed. Best for autonomous, recurring work.
• Desktop tasks: Runs on your machine. Needs computer on. Best when you need local file access.
• /loop: Session-scoped polling. Quick, lightweight, but doesn't persist across restarts.

Cloud tasks have a minimum interval of 1 hour. Desktop tasks can run every minute. Choose based on whether you need your machine involved.

What About Security?

Claude only pushes to claude/-prefixed branches by default. This prevents accidental modifications to protected branches. You can enable unrestricted branch pushes for specific repos if needed.

Connectors are opt-in per task. You control exactly which external services each scheduled task can access.

The Bottom Line

This is the shift from "AI as a tool I use" to "AI as a teammate that works independently." Scheduled tasks aren't just convenient—they fundamentally change what's possible.

Your AI can now maintain codebases, monitor systems, and handle routine work without your involvement. Set it up once, review the results, iterate on the prompt. Over time, you offload more and more.

The future of work isn't about doing more tasks. It's about designing systems that do them for you.

Claude just made that a lot easier.