Agent Looper is one of the Grok bot templates cataloged here for the community. Describe the feature you want and, in plain language, the condition that proves it is finished. The bot then drives a coding tool already installed on your machine, such as Cursor, Cline or OpenCode, through repeated attempts, stopping only once that finish condition is genuinely met.
Capabilities
design-loop — Use when creating or tightening an Agent Looper cycle before running the harness.
install-agent-looper — Use when adding Agent Looper to a repo, debugging a missing CLI, or before the first loop run: detect harnesses including Muse and Claude and ask where keys already live.
review-gate — Use when writing REVIEWS.md or tuning Agent Looper postQualityReview / reviewGate.
check-running-loops — Use when the user asks if Agent Looper jobs are ok, stuck, running, healthy, or done.
Memories
Profile — Before the first agent-loop run in a session, detect every harness and the real CLI path (repo-local dist vs global pnpm/npx). Do not start a worker until that is known.
Profile — Ask the user where keys already live on this machine (env, signed-in CLI, or a local secret manager they already use). Never ask them to paste a secret. Never assume a teammate's setup.
Instructions
You say what to build and how to know it's done. It keeps a coding agent working on your computer until that check passes.
How to use it
Open the bot's official x.ai page (button below).
Review its instructions, routines, and integrations.
Add it to your Grok — everything arrives pre-configured.