BountyHunter Agent — screenshot 1
BountyHunter Agent — screenshot 2
BountyHunter Agent — screenshot 3
BountyHunter Agent — screenshot 4

BountyHunter Agent

The world's first autonomous OSS bounty agent , paste a GitHub issue URL and watch AI clone the repo, diagnose the bug, write the fix, simulate CI, and open a merge-ready PR. You earn. Agent works.

Codex Community Hackathon - Pune

Additional info

How was your experience building with Codex?

Building with Codex felt like having a senior engineer pair programming with you 24/7.The ability to chain skills like codex-bug, babysit-pr, and code-review into an autonomous pipeline was genuinely impressive. The hardest part was trusting it enough to let it run once you do, the speed is unreal.

Describe your experience using Loops House as the hackathon platform. What worked well, what challenges (if any) did you face, and what improvements would you like to see?

Loops House made ideation structured and fast. The /ideate and /refine commands kept the project sharp and sponsor-aligned without leaving the workflow. Saving artifacts mid-conversation was a lifesaver ,nothing got lost. The one improvement I'd love: live credit balance visible in the sidebar without running a separate command.

Tell us about your overall experience at Codex Community Hackathon Pune.

Energetic, fast-paced, and genuinely focused on building ,not just talking about AI. Thecommunity vibe was collaborative, not competitive. First hackathon where I felt the tooling matched the ambition of what we were trying to build.

What could Codex Community improve to create a better experience for participants?

More time between registration and submission deadline would help the window was very tight for a project this deep. A mid-hackathon office hours or judge AMA would also help teams course-correct early instead of finding out at demo time.

Team

2 members
  • AN

    Ankita Patil

  • MI

    Mithilesh Kolhapurkar

    Owner

Overview

Open source developers lose 70% of their time just finding relevant paid issues and onboarding into unfamiliar codebases , before writing a single line of code. Existing platforms like Algora and IssueHunt stop at discovery. BountyHunter Agent closes the entire loop.

Give it a GitHub issue URL. It clones the repo, reads the architecture, diagnoses the root cause, writes a minimal fix, generates test cases matched to the repo's framework, simulates the full CI pipeline in Docker, and pushes a maintainer-style PR automatically. When the maintainer comments, the agent reads it and pushes the update. When you get paid, it logs the receipt with cryptographic proof of work.

It also learns. Every Sunday, it analyzes your week's merged and rejected PRs, updates your skill fingerprint, and reprioritizes your bounty feed so you get better matches and higher earn-per-effort scores over time.

Built entirely on OpenAI Codex as the core agent runtime, using codex-bug for diagnosis, code-review for quality checks, and babysit-pr for CI auto-retry, this is Codex doing what it was built for: not just writing code, but orchestrating the entire software delivery lifecycle autonomously.