ehmo

ehmo / autoresearch

Public

Autonomous codebase improvement

19
0
100% credibility
Found Mar 17, 2026 at 19 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Shell
AI Summary

Autoresearch uses AI teams to autonomously find issues, fix bugs, and simplify code in a project by running improvement cycles on a separate branch.

How It Works

1
🔍 Discover autoresearch

You hear about a clever tool that lets smart helpers improve your code automatically while you relax.

2
🛠️ Set it up quickly

You run a simple setup script to add this magic to your AI coding companion.

3
📁 Point to your project

You tell it the folder with your code using an easy chat command, and it gets ready to help.

4
🚀 Watch teams improve code

Three separate smart teams take turns spotting issues, fixing bugs, and simplifying things on a safe test version of your work.

5
📊 Check in anytime

You glance at updates to see progress, resume if needed, and decide when it's perfect.

🎉 Celebrate better code

Your project emerges cleaner, faster, with fewer problems – just merge it into your main work and enjoy.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 19 to 19 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 autoresearch?

Autoresearch runs autonomous AI teams on your codebase to find bugs, fix them, and simplify code—all on a Git feature branch that never touches main. Inspired by Karpathy's auto research and forks like pi-autoresearch, it loops through red team analysis, green team fixes with test verification, and refactor simplifications until improvements slow. Built in Shell, it plugs into Claude Code via slash commands like /autoresearch ~/project or /autoresearch resume, auto-detecting test commands for stacks like Go, Node, Rust, and Python.

Why is it gaining traction?

It stands out with info barriers between teams—fixers don't see how issues were spotted—for cleaner, unbiased changes, unlike single-agent autonomous coders on GitHub. Users get detailed logs, TSV results, and deferred ideas, plus config for includes/excludes and cycle limits, making autonomous codebase improvement hands-off and verifiable. Devs dig the Karpathy-style "set it and forget it" hook, generalized beyond ML to any tested repo.

Who should use this?

Backend engineers maintaining mid-sized Go or Rust services with solid test coverage, seeking autonomous GitHub Copilot-style cleanup without manual reviews. Teams refactoring legacy Node/Python monorepos via include/exclude rules, or solo devs experimenting with auto research on open-source projects.

Verdict

Try it on a test repo if you have Claude Code and tests—docs are solid, install is one script, but 19 stars and 1.0% credibility signal early-stage alpha; expect tweaks for your stack. Worth the 5-minute setup for autonomous codebase gains, but skip without tests.

(178 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.