Free guide · for everyone who commented "LIBRARY" · reading time ~10 min · setup for your first loop: ~2 minutes

📄 The fully illustrated version — with the loop-cycle diagram, the command cards, and Loop Library screenshots — is the downloadable PDF in the Guide Library.

FIND THE VISUAL PDF HERE -

The-5-AI-Loops-I-Run-Every-Day.pdf

The short version

What a loop actually is

A loop is a task you hand to an AI agent that it repeats on its own until a stop condition is met. You're not re-prompting it every time — you set it once and it runs, checks its own work, and keeps going (or stops) based on a rule you gave it. That's what turns AI from a thing you operate into a thing that operates for you.

The best public catalogue of these is the Loop Library from Forward Future — 70 practical agent loops, each with clear checks and a stopping condition:

https://signals.forwardfuture.com/loop-library/

The Loop Library — 70 copy-paste agent loops with built-in checks and stop conditions.

The 3 commands that power every loop

In Claude Code, almost every loop is one of three commands pointed at the right job. That's the entire toolkit.

Command Runs Best for
/loop [interval] [task] On a timer, on your machine Frequent, lightweight checks while you work
/schedule [when]: [task] In the cloud — even when your computer is off Daily/weekly jobs that must run regardless
/goal [condition] Until a measurable target is hit Open-ended work with a clear finish line

The 5 I run every day

1. The nightly log

What it does: every night it writes up what changed across my projects — what shipped, what moved, what's still open — into a running file. I wake up to a diary of progress instead of trying to remember where I left off. Why it earns daily use: context is the thing you lose overnight; this hands it back every morning for free.