congwa

congwa / codex-autoresearch

Public

codex 永动机，避免codex执行一个长任务总是自动停止的问题

43
3
69% credibility
Found Mar 26, 2026 at 43 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Shell
AI Summary

A simple script that keeps an AI coding assistant persistently working on long development tasks by repeatedly resuming until a precise completion confirmation is received.

How It Works

1
📰 Discover the task keeper

You hear about a handy tool that lets your AI coding helper tackle big jobs without stopping halfway.

2
📝 Write your to-do list

Jot down simple instructions for what you want done, like fixing bugs or updating guides, in a plain note.

3
📁 Choose your project spot

Point the tool to the folder with your project so the AI knows where to work.

4
🚀 Launch the keeper

Hit go, and your AI buddy starts chipping away at the job while the keeper watches over it.

5
⏳ Let it run unattended

Walk away—the tool keeps nudging the AI to continue round after round until the job's truly done.

6
🔄 Restart if needed

If something interrupts, just restart the keeper and it picks right up where it left off.

🎉 Job complete!

The AI gives the all-clear signal, your project is fixed up, and everything works great.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 43 to 43 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is codex-autoresearch?

Codex-autoresearch is a Bash script that keeps Codex CLI sessions alive for long-running tasks, kicking off with `codex exec` and looping `codex exec resume` until a strict two-line completion protocol confirms the job's done—no trusting vague "I'm finished" claims. It solves Codex halting midway on extended autoresearch jobs due to timeouts or disconnects, preserving context across resumes via session IDs or `--last`. Feed it a prompt file or stdin, tweak via env vars like WORKDIR for repo targeting or STATE_DIR for state persistence.

Why is it gaining traction?

It stands out with bulletproof completion via randomized tokens, dodging model hallucinations, plus seamless recovery from interruptions using saved events and session IDs—perfect for chaining into cron or tmux. Developers dig the GitHub integration hooks, like running Codex on issues or releases without babysitting, and options for models like codex max or mini via `-m`. No more context drift or repeated startups; just reliable, unattended progress on codex github copilot-style workflows.

Who should use this?

Repo maintainers automating TODO fixes or test completions in monorepos. DevOps folks scripting Codex for GitHub actions on open issues. AI tinkerers chaining codex cli with pipelines for ongoing autoresearch, like doc updates across codex_home projects.

Verdict

Grab it for niche Codex long-hauls if you already use the CLI—docs are solid, setup's a one-liner—but with 43 stars and 0.7% credibility score, it's early-stage; test in a sandbox first. Pairs well with codex config toml for profiles, but lacks broad polish for production pipelines.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.