agent-autoloop
KKKKhazix/agent-autoloopReviewed by AI
Checked by RepoPicks Radar on
What is agent-autoloop?
A background service for people who already use terminal AI assistants such as Claude Code or Codex. You write rules in a JSON file (for example: every morning at 9, or when a new GitHub issue arrives) and it starts the assistant, follows the job through, and sends you a card on desktop, Feishu, Slack or Discord when it needs your choice. While it waits, nothing is running. It works on macOS and Linux and you need Node 24, Git, curl and an assistant you have already logged into. Windows is not supported yet.
How easy is it to install agent-autoloop?
D · terminal, several tools or a specific graphics card · macOS
- macOS D
- Linux D
On macOS and Linux it is all terminal: install Node 24, Git and curl, run an install script, log into a command-line AI assistant, then write rules by hand in a JSON file. There is no grade for Windows because the README says it does not run there yet.
Coming soon: install in 1 click with the RepoPicks app.
agent-autoloop on GitHub
- Stars
- 4,420on Nov 10, 2026
- Downloads
- —all releases
- Growth
- +4,420stars in 7 days
Stars over time
Facts
- Runs on
- macOS, Linux
- You need to install
- Node.js, Other tools
- To build from source
- Node.js
- Disk space
- 100 MB–500 MB · estimate
- Accounts or keys
- None
- Cost
- Free, API costs extra · The software is MIT and free. The AI work runs on your own assistant account, so you pay the model bill; the README says that for heavy automation the model providers prefer an API key over a personal subscription.
- License
- MIT (Permissive)
- Last change on GitHub
- Nov 10, 2026
agent-autoloop risk signals (5)
- Attentioninstall script piped from internet
INSTALL.md and README.md give the install command as `curl -fsSL https://raw.githubusercontent.com/KKKKhazix/agent-autoloop/main/install.sh | sh`, which runs a downloaded script straight away; install.sh then clones the repository into ~/.agent-autoloop/app and installs a background service with `"$PROGRAM" service install`.
AI review
- Attentionai runs commands unattended
The whole point is unattended work: README.md says the hub decides "交给哪个 Agent、去哪个目录干" from your rules, and the delivery pipeline "审代码 → 等 CI → 合并 PR → 等部署" merges pull requests on its own after a `deliver` label. The project says it asks you first for things like deleting data, restarting services or spending money, which is a promise that only running it can confirm.
AI review
- Infovery new project
The repository was created on 2026-11-03 and had 2404 stars and a single release (v1.0.0, 2026-11-03) when it was checked on 2026-11-04, so there is no track record yet.
AI review
- Infofeatures not tested by author
README.md states the four webhook templates were only checked against a local mock server, the GitLab webhook was never connected to a real GitLab, and the listed command lines for other assistants (DeepSeek Harness, pi, OpenClaw, Hermes, Cursor, CodeBuddy) were copied from their docs and "我们还没有逐个跑过" (not tried one by one).
AI review
- Infodocs mostly in chinese
The main README.md, INSTALL.md and the terminal messages inside install.sh are in Chinese; there is an English README.en.md, but the install notes and examples you have to follow are Chinese.
AI review
Evidence is shown in its original language.
What Radar could not check about agent-autoloop
- Whether the install script and the background service really do only what the documents describe on your machine
- Whether it works end to end on macOS and Linux with your own assistant login
- How much the model usage costs once rules are running every few minutes
- Whether the promise of asking before deleting data, restarting services or sending messages holds in practice
- Whether the announced Windows version arrives
- Real disk use as the append-only log and database grow over time