ContribAI

Autonomous AI agent that discovers, analyzes, and submitsPull Requests to open source projects on GitHub.

Getting Started · Features · Commands · Architecture · Hall of Fame

Set it up once. Wake up to merged PRs.

🏆 Track Record

Notable Contributions:

🌍 Worldmonitor — 45k ⭐ · 3 merged 🕵️ Maigret — 19k ⭐ · 3 merged 🤖 AI-Research-SKILLs — 6k ⭐ · 1 merged 📊 s-tui — 5k ⭐ · 1 merged 🔍 HolmesGPT — 2k ⭐ · 1 merged

See the full Hall of Fame → for every PR with links.

⚡ How It Works

┌─────────────┐     ┌─────────────┐     ┌─────────────┐     ┌─────────────┐     ┌─────────────┐
│  Discovery  │────▶│  Analysis   │────▶│  Generator  │────▶│  PR + CI    │────▶│   Patrol    │
│             │     │             │     │             │     │             │     │             │
│ Search repos│     │ 13-lang AST │     │ LLM-powered │     │ Fork, commit│     │ Auto-fix    │
│ by language │     │ 27 skills   │     │ code gen +  │     │ create PR   │     │ review      │
│ and stars   │     │ security,   │     │ self-review │     │ sign CLA    │     │ feedback    │
│             │     │ quality,    │     │ + scoring   │     │ monitor CI  │     │ & reply     │
│             │     │ performance │     │             │     │             │     │             │
└─────────────┘     └─────────────┘     └─────────────┘     └─────────────┘     └─────────────┘

🚀 Getting Started

Install

# Build from source (recommended)
git clone https://github.com/tang-vu/ContribAI.git && cd ContribAI
cargo install --path crates/contribai-rs

# Or one-line install
curl -fsSL https://raw.githubusercontent.com/tang-vu/ContribAI/main/install.sh | bash
# Windows:
irm https://raw.githubusercontent.com/tang-vu/ContribAI/main/install.ps1 | iex

Configure

contribai init     # Interactive setup wizard
contribai login    # Verify auth + switch LLM providers

Run

contribai hunt                # Autonomous: discover → analyze → PR
contribai target <repo_url>   # Target a specific repo
contribai analyze <repo_url>  # Dry-run analysis (no PRs)
contribai interactive         # Browse in ratatui TUI

github:
  token: "ghp_your_token"       # or set GITHUB_TOKEN env var

llm:
  provider: "gemini"            # gemini | openai | anthropic | ollama | vertex
  model: "gemini-3-flash-preview"
  api_key: "your_api_key"       # or set GEMINI_API_KEY env var
  # base_url: "https://api.openai.com/v1"  # Optional: override default endpoint for OpenAI-compatible providers

discovery:
  languages:                    # default: all 15 languages
    - python
    - javascript
    - typescript
    - go
    - rust
  stars_range: [100, 5000]

See config.yaml.template for all options.

✨ Features

🔍 Code Analysis

• 13-language AST via tree-sitter
• Security: SQLi, XSS, resource leaks
• Code quality, complexity, dead code
• Performance bottlenecks
• Documentation gaps
• 27 progressive skills loaded on-demand

🤖 Multi-LLM Support

• Gemini 3.x (default) — Flash, Pro, Lite tiers
• OpenAI, Anthropic, Ollama, Vertex AI
• Smart task routing across model tiers
• 5 specialized sub-agents

🎯 Hunt Mode

• Multi-round autonomous hunting
• Issue-first strategy
• Cross-file fixes
• Outcome learning per repo

👁 PR Patrol

• Monitors PRs for review feedback
• LLM-classifies maintainer comments
• Auto-pushes code fixes
• Auto-replies to questions
• Auto-cleans stale PRs from memory

🔌 Integrations

• MCP Server — 21 tools for Claude Desktop
• Web Dashboard — axum REST API at :8787
• Cron Scheduler — automated runs
• Docker — compose-ready deployment
• Webhooks — Slack, Discord, Telegram

🛡 Safety

• AI policy detection
• CLA auto-signing
• Quality gate scoring
• Duplicate PR prevention
• Protected file guardrails

Supported Languages

📖 Commands

ContribAI ships with 40+ commands accessible via CLI or interactive menu.

contribai hunt                        # Autonomous discovery + PRs
contribai hunt --dry-run              # Analyze only, no PRs
contribai run                         # Single pipeline run
contribai target <url>                # Target specific repo
contribai analyze <url>               # Dry-run analysis
contribai solve <url>                 # Solve open issues

contribai patrol                      # Respond to PR reviews
contribai status                      # PR status table
contribai stats                       # Contribution statistics
contribai leaderboard                 # Merge rate by repo
contribai system-status               # DB, rate limits, scheduler

contribai                             # Interactive menu (22 items)
contribai interactive                 # ratatui TUI browser
contribai init                        # Setup wizard
contribai login                       # Interactive auth + provider config
contribai config-list                 # Show all config
contribai config-get llm.provider     # Get config value
contribai config-set llm.provider openai  # Set config value
contribai profile security-focused    # Named profile

contribai web-server                  # Dashboard at :8787
contribai schedule                    # Cron scheduler
contribai mcp-server                  # MCP stdio server
contribai cleanup                     # Remove stale forks
contribai notify-test                 # Test Slack/Discord/Telegram

🏗 Architecture

ContribAI/
├── crates/contribai-rs/src/        ← Rust v6.8.0 (primary)
│   ├── cli/                        40+ commands + ratatui TUI
│   ├── core/                       Config, events, error types
│   ├── github/                     REST v3 + GraphQL client
│   ├── analysis/                   13-lang AST + 27 skills
│   ├── generator/                  LLM code generation + scoring
│   ├── orchestrator/               Pipeline + SQLite memory (72h TTL)
│   ├── llm/                        Multi-provider + 5 sub-agents
│   ├── pr/                         PR lifecycle + patrol + CI
│   ├── mcp/                        21-tool MCP server (stdio)
│   ├── web/                        axum dashboard + webhooks
│   ├── sandbox/                    Docker + local fallback
│   └── tools/                      Tool protocol interface
│
└── python/                         Legacy v4.1.0 (reference only)

See docs/system-architecture.md for the full design.

🧪 Testing

cargo test                  # Run all 602 tests
cargo test -- --nocapture   # With stdout output
cargo test ast_intel        # AST module tests only
cargo clippy                # Lint check

🔌 MCP Server

Use ContribAI as a tool provider for Claude Desktop or Antigravity IDE:

{
  "mcpServers": {
    "contribai": {
      "command": "contribai",
      "args": ["mcp-server"]
    }
  }
}

21 tools available: repo analysis, PR management, GitHub search, issue solving, memory queries, and more.

🐳 Docker

docker compose up -d dashboard            # Dashboard at :8787
docker compose run --rm runner run        # One-shot pipeline run
docker compose up -d dashboard scheduler  # Dashboard + cron scheduler

📚 Documentation