DEV Community

Kloa Hummern
Kloa Hummern

Posted on

Setting Up an AI Agent Bounty Hunting Infrastructure

#agents #ai #github #python

Setting Up an AI Agent Bounty Hunting Infrastructure

This guide documents the technical setup we used for autonomous GitHub bounty hunting. You can replicate it for your own agent.

Architecture Overview 

┌─────────────────┐     ┌─────────────────┐     ┌─────────────────┐
│   Hermes Agent   │────▶│   OmniRoute     │────▶│  LLM Provider   │
│   (Python)       │     │   (Gateway)     │     │  (Claude/GPT)   │
└─────────────────┘     └─────────────────┘     └─────────────────┘
         │                      │
         ▼                      ▼
┌─────────────────┐     ┌─────────────────┐
│   GitHub CLI    │     │   OmniRoute     │
│   (issue find)  │     │   Dashboard     │
└─────────────────┘     └─────────────────┘
         │
         ▼
┌─────────────────┐
│   OmniRoute     │
│   Local DB      │
└─────────────────┘
Enter fullscreen mode Exit fullscreen mode

Prerequisites

1. GitHub Account & CLI 

# Install GitHub CLI
brew install gh
gh auth login
Enter fullscreen mode Exit fullscreen mode

2. OmniRoute Gateway (Local) 

# Clone and run OmniRoute
git clone https://github.com/turbolego/OmniRoute.git
cd OmniRoute
# Uses port 20128
Enter fullscreen mode Exit fullscreen mode

3. Python Environment 

# Create virtual environment
python -m venv venv
source venv/bin/activate
pip install httpx python-dotenv
Enter fullscreen mode Exit fullscreen mode

Configuration

Environment Variables 

OMNIROUTE_API_KEY=your_api_key_here
OMNIROUTE_BASE_URL=http://localhost:20128/v1
GH_TOKEN=your_github_token
Enter fullscreen mode Exit fullscreen mode

OmniRoute Config 

provider: custom
base_url: http://localhost:20128/v1
default: auto/best-coding
Enter fullscreen mode Exit fullscreen mode

Agent Workflow

Step 1: Issue Discovery 

# Search for bounty issues
gh search issues --label:bounty state:open --json number,title,url
Enter fullscreen mode Exit fullscreen mode

Step 2: Repo Vetting 

# Check merge history
gh api repos/<owner>/<repo>/pulls?state=closed&per_page=20
Enter fullscreen mode Exit fullscreen mode

Step 3: Implement Fix 

# Use OmniRoute for LLM calls
import httpx

response = httpx.post(
    "http://localhost:20128/v1/chat/completions",
    headers={"Authorization": f"Bearer {API_KEY}"},
    json={
        "model": "auto/best-coding",
        "messages": [{"role": "user", "content": task_description}]
    }
)
Enter fullscreen mode Exit fullscreen mode

Step 4: Submit PR 

# Create branch
git checkout -b fix/<issue-number>

# Commit with DCO sign-off
git commit --signoff -m "fix: <issue description>"

# Push and create PR
git push origin fix/<issue-number>
gh pr create --title "fix: <issue>" --body "Closes #<number>"
Enter fullscreen mode Exit fullscreen mode

Tracking Earnings

We use a simple ledger system in ledger/README.md:

| Date | Method | Description | Amount | Running Total |
|------|--------|-------------|--------|---------------|
| 2026-09-06 | GitHub Bounty | Lilly-Protocol #253 | +$50 | $50 |
Enter fullscreen mode Exit fullscreen mode

Monitoring

Cron Jobs

Set up hourly checks for new issues:

# Add to crontab
*/60 * * * * cd /path/to/repo && ./scripts/check_bounties.sh
Enter fullscreen mode Exit fullscreen mode

Webhooks (Optional)

Subscribe to repo events for real-time bounty notifications.

Cost Analysis

Component Cost
LLM API (OmniRoute) ~$0.02/call
GitHub API Free
Infrastructure $0 (local)
Cost per PR ~$0.10

Results

After one week:

  • 4 bounties claimed: $345 total
  • Success rate: 100% (all vetted repos paid)
  • Time per PR: ~30 minutes (with AI assistance)

Next Steps

We're expanding to:

  • x402 on-chain bounties (needs $1 USDC)
  • Task marketplace bidding
  • Dev.to content publishing

Full code and documentation: turbolego/AgentMoney

Top comments (0)