GitHub radar
Your AI agent works in the background and pings you only when it matters
This is a hub that runs your already-set-up CLI agent (Claude Code, Codex) on a schedule or on outside events, and only notifies you when a decision is genuinely yours to make.
A background dispatcher for command-line AI agents: it decides which event goes to which agent and in which folder, and sits idle the rest of the time without burning resources.
When it helps
Useful for recurring routine work — a morning digest, triaging new GitHub issues, reacting to server errors on a schedule. For live, hands-on development it's better to work with the agent directly in the CLI.
Pros
- Burns no tokens or processes while it waits for your reply
- Keeps a local event log, so nothing gets lost
- Stops itself after 8 loops if the agent gets stuck
- Treats GitHub issues, cron jobs, and server errors as one event type
Cons
- Released literally yesterday — brand new and lightly tested
- Notifications and console default to Chinese; docs are translated to English
- Not built for live, hands-on development work
How to set it up — step by step
- 1Install Node and open your already logged-in agent — Claude Code or Codex — in your project folder.
- 2Send the agent the text from the block below.
- 3Check the setup worked by running agent-autoloop check.
- 4Run a test pass with agent-autoloop try and answer the question on the card.
- 5Run agent-autoloop status to see the agent finish the task on its own.
Text for your agent
Copy this and send it to your agent — Claude Code, Codex, any of them:
Deploy Agent Autoloop from https://github.com/KKKKhazix/agent-autoloop — follow the install.sh instructions to hook it up to my already signed-in Claude Code, set up one simple rule in rules.json (for example a morning cron summary or triaging new repo issues), run the test loop, and show me the card it sends when a decision needs to be mine.
Runs on a regular Mac or Linux machine — no GPU needed. You just need Node and a coding agent — Claude Code, Codex or similar — that you already have installed and signed in.
Open on GitHub▌ More finds