This GitHub repository offers the "loop-engineer-template," a starter kit for building autonomous AI agents designed to wake on triggers, perform work, verify outcomes, and log learnings into a shared, file-based memory. The core concept is "compounding" work, where multiple specialized loops (e.g., support, product, SEO) interact with a common knowledge base, allowing insights from one to inform and accelerate others. The template provides a structured knowledge base, a codebase harness, and necessary skills to enable agents to autonomously run, test, and ship code changes. It aims to shift development from task-by-task prompting to designing self-sustaining agent loops.
It's the productized version of the setup my team runs in production, and what I teach at AI Builder Club. What's a loop engineer? The shift: you stop prompting a coding agent task-by-task, and start designing loops. A loop is an agent that wakes up on a trigger (a cron, a webhook, an incident, another agent), does some investigation and work, and writes what it found and did into a shared, file-based memory. Next run it reads that memory and keeps going.