What are Claude Loops?

Claude Loops are recurring workflows in Claude Code that automatically run a prompt or slash command on a defined interval. You invoke them with the /loop command, e.g. /loop 5m /check-deploy, and Claude repeats the task on schedule until you stop it.

How do Claude Loops work?

Claude Loops use two modes. Fixed-interval mode (/loop 5m /task) runs your task every N minutes. Dynamic mode (/loop /task without an interval) lets Claude self-pace by calling ScheduleWakeup with a delay it chooses based on what it is waiting for. Both modes survive context compression and continue until you cancel.

How do I create my first Claude Loop?

Open Claude Code in your project and type /loop followed by an optional interval and your prompt or slash command. Example: /loop 10m check the build status and report failures. Claude will run the task immediately, then again every 10 minutes until you stop the session or type /loop stop.

Are Claude Loops the same as cron jobs?

No. Cron jobs run shell commands on a server clock with no intelligence between executions. Claude Loops run an AI agent that can read context, make decisions, and adjust its own pacing. Loops are session-bound by default, while CronCreate-based agents run independently in the cloud.

Automate recurring tasks with Claude Code Loops & Generator

Claude Loops let you run any prompt or slash command on a repeating interval. Poll a CI build, babysit a deploy, watch a Slack thread - all from a single command. Beginner-friendly, no scripting required.

⚡ Generate a Loop Now → 5-Min Starter Guide

Why Claude Loops

Recurring AI workflows, without writing a single line of code

Loops bring the intelligence of Claude Code to repeating tasks. Anything you do once, you can now do on schedule - with the same reasoning behind every run.

One command, infinite reruns

Type /loop with your task. Claude handles the scheduling, context, and stopping - you stay in flow.

Smart, self-pacing mode

Drop the interval and Claude picks a wake-up time based on what it is waiting for. Polls fast when state changes, sleeps long when idle.

Works with any slash command

Wrap your favorite skill - /review, /verify, /code-review - in a loop and let it run on a schedule.

Safe and cancellable

Loops are session-bound. Stop them anytime with /loop stop or by ending the session. No background processes left behind.

Real-world use cases

Monitor PRs, babysit deploys, poll APIs, watch logs, summarize Slack channels. If you can describe it, you can loop it.

Local or cloud

Run loops locally in your terminal or schedule them as cloud agents with /schedule for true unattended automation.

⚡ Try it now - no signup

Build your /loop command right here

Tell the generator what you want to automate. It writes a ready-to-paste /loop command with the right mode, filters, and stop condition baked in.

1. Build your loop

Fill in what you need. The command updates as you type.

Tip: A great loop has three ingredients - a clear task, an "only tell me when…" filter, and a stop condition. The form below builds all three for you.

Start with

What should Claude do each iteration? Write it like you would say it out loud. e.g. "check the latest CI build", "tail the error log".

How often? Let Claude self-pace (recommended) or pick a fixed interval.

How loud should it be? Silence is golden. Most loops should only interrupt you on change.

When should it stop? Optional but recommended. e.g. "when the build passes", "once production is live". Leave blank if you want it to run until you halt it.

Response style

Open Generator on its own page →

How To Use

Get your first loop running in 4 steps

You do not need to install anything new. If you have Claude Code, you already have loops.

Open Claude Code

Launch Claude Code in any project directory. Loops work in CLI, IDE extensions, and the desktop app.

Type the /loop command

Enter /loop followed by an optional interval (like 5m) and your task or another slash command.

Watch it run

Claude executes the task immediately, then again on schedule. Each run gets fresh context and reports back.

Stop when done

Type /loop stop or close the session. The loop ends cleanly - no stray processes.

Full Step-by-Step Guide →

Real Examples

What people actually loop

From CI polling to morning standup digests - here are the loops developers run every day.

Watch a CI build

Poll a long-running build, summarize failures, and alert you when it finishes.

/loop 3m check CI status

Babysit open PRs

Loop a review skill that scans new comments and pings you on blockers.

/loop 15m /babysit-prs

Auto-summarize logs

Tail an error log and surface new patterns every few minutes.

/loop 5m scan logs for errors

Inbox triage

Loop a Gmail skill that classifies new threads and drafts replies.

/loop 30m triage inbox

Track metrics

Fetch dashboard numbers, diff them against the last run, and ping if they spike.

/loop 10m metric check

Daily digests

Run a morning briefing skill once a day with /schedule.

/schedule 9am morning brief

See 20+ Loop Examples →

Ready to build your first Claude Loop?

Follow our 5-minute starter guide and you will have a working, real-world loop running before your coffee gets cold.

Start the Guide →

Quick Answers

Claude Loops, in plain English

What is a Claude Loop in one sentence?

A Claude Loop is a recurring prompt or slash command that Claude Code runs on a schedule you choose - fixed interval, or self-paced by Claude itself.

Do I need an API key or extra tools?

No. Loops are built into Claude Code. If you can run Claude Code, you can run loops - no extra setup, no separate billing, no infrastructure to manage.

Will a loop drain my context window?

No. Claude Code automatically summarizes prior messages as the conversation approaches context limits. Loops are designed to keep running across compressions - you do not have to wrap up early.

How is a loop different from /schedule?

/loop runs inside your active session. /schedule creates a cloud agent that runs without you. Use loops while you are working; use schedules for unattended automation.

Read All FAQs →