GitHub Action · $15/month · BYOK

AI agents that find
real bugs in your PRs

A fleet of AI agents reviews every pull request, debates each finding, and only reports bugs that survive consensus verification.

Get PR Hunt Read the docs
pr-hunt output
$ pr-hunt --pr 847 --fleet-size 5
[14s] Spawned bug_finder_early agent: agent-a1b2
[14s] Spawned bug_finder_early agent: agent-c3d4
[38s] Bug submitted: off-by-one in pagination
[52s] Bug submitted: unchecked null in auth flow
[71s] Verification: 5/5 confirmed · 0 refuted
CONFIRMED [high] unchecked null in auth flow
File: src/middleware/auth.ts:42
CONFIRMED [medium] off-by-one in pagination
File: src/api/list.ts:118
✓ 2 bugs confirmed, 0 false positives

How it works

Three steps. Zero infrastructure.

01

Subscribe

$15/month. Cancel anytime. You get a license key emailed instantly.

02

Add the workflow

Drop a YAML file into your repo. Add your license key and AI API key as secrets.

03

Ship with confidence

Every PR gets reviewed by AI agents. Real bugs get flagged. False positives get voted out.

.github/workflows/pr-hunt.yml 
name: PR Hunt
on:
  pull_request:
    types: [opened, synchronize]

jobs:
  hunt:
    runs-on: ubuntu-latest
    steps:
      - uses: pr-hunt/pr-hunt@v1
        with:
          license-key: ${{ secrets.PR_HUNT_LICENSE_KEY }}
          openai-api-key: ${{ secrets.OPENAI_API_KEY }}
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

Features

Not another linter.

PR Hunt finds logic bugs, race conditions, and security issues that static analysis misses.

Multi-Agent Consensus

Multiple AI agents independently review your code. Bugs are only reported after surviving adversarial verification — 5 confirmations required, skeptical verifiers trying to disprove each finding.

Bring Your Own Keys

Use your own OpenAI or Anthropic API keys. You control the cost. No data leaves your GitHub Actions runner except API calls you authorize.

Native GitHub Action

Runs entirely inside GitHub Actions. No external servers, no webhooks to configure, no OAuth flows. Add a YAML file and you're done.

Zero Hosting Cost

Everything runs on GitHub's infrastructure. No servers to maintain, no databases to manage. Runs on your own compute.

Inline PR Comments

Confirmed bugs are posted directly as inline comments on the exact lines that contain issues. Review the AI's findings alongside your normal code review workflow.

Dual Engine Support

Choose between OpenAI Codex and Anthropic Claude as your agent engine. Use whichever model family you prefer, or switch between them per-repo.

Stop shipping bugs.

$15/month. Cancel anytime. No per-seat fees.
Your code stays on GitHub's servers.

View Pricing