Agenthunter drops a pack of AI agents into your app. They click around, catch what's broken, write the patch, and prove it works before anyone on your team lifts a finger.
Every hunter has one job and does it well. Plug in whichever AI model you like for each one: cheap and fast for the grunt work, big brain for the final call.
Walks every screen of your app like a real user. Taps, types, scrolls, and sniffs out anything that looks off, down to hidden console errors.
sees the screenSorts real bugs from false alarms. Every confirmed catch gets a clean report with screenshots and exact steps to reproduce it.
no noise, only factsWrites the code fix in a safe sandbox copy of your repo. The Tracker and the Warden then re-check it in the live app before it ships.
patches, then proves itStart a hunt and walk away. Every time new code lands, the pack goes back in. No new code? It stays quiet and costs almost nothing.
Then it waits for the next commit and does it all again. Auto-fixing and GitHub posting stay off until you switch them on.
If it runs on your computer, the pack can hunt in it. Tell it how to log in, in plain English, and it takes it from there.
Your model, your choice. Run it on Claude Code, Codex, Kimi, or any AI you already pay for, or plug in an API key from OpenAI, Anthropic, OpenRouter, and more.
The meme coin of the pack. Bugs get hunted, the community gets the memes.
All you need is Node 22+, git, and a way to start your app. Setup sniffs out your project and the AI tools you already have.
Read the guide# point the pack at your repo $ npx agenthunter init # let them loose $ npx agenthunter patrol # watch the hunt live $ npx agenthunter dashboard