Lightweight task automation middleware that lets command-line AI agents run in the background.
Run command-line AI agents in the background, waking you only when decisions are required.
Agent Autoloop is a zero-dependency local middleware designed to schedule tasks, manage agent events, and push notification cards to channels like Feishu or desktop alerts. It is built for developers and technical professionals who use command-line AI agents like Claude Code or Codex and want them to run unattended.
The tool acts as a lightweight event-driven runner that parses incoming tasks, assigns them to CLI agents based on your rules, and exits to consume zero tokens or CPU while waiting for human input. When a decision is needed, it sends an interactive card to your chosen communication channel, resuming the exact session once you reply.
It suits technical professionals and developers who regularly rely on CLI-based coding agents and want to automate recurring workflows like morning briefs, issue triage, and code delivery. It does not suit non-technical users who require a graphical web interface or those unwilling to configure JSON rules.
Processes exit entirely while waiting for your input, consuming zero tokens and background resources.
Define custom triggers via a simple JSON rules file mapping events and crons to specific agents.
Push interactive decision cards to Feishu, desktop notifications, Slack, Discord, or DingTalk.
Prevents runaway agent loops by automatically halting recursion after eight nested event triggers.
The software is free under its open-source license, and the only cost is running it on your own hardware along with the API fees consumed by your underlying AI agents.
873 open remote AI-training roles, updated November 15, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
These reviewed AI tools do the same job instantly — from our scored directory.
No, it runs entirely on your local Windows, Mac, or Linux machine as a lightweight background service.
The agent exits after each run and sends an interactive decision card to your notification channel; your reply resumes the exact same session.
It includes built-in support for Claude Code, Codex, ACP, DeepSeek Harness, and standard CLI commands.
See Zekai first in Google