AI-Powered GitHub PR Bot

🤖 Intelligent pull request analysis, review, and auto-approval using Claude AI

Transform your pull request workflow with AI-powered automation that provides intelligent code reviews, generates comprehensive descriptions, and safely auto-approves quality changes.

✨ Why Use This Bot?

• 🔍 Catch Issues Early - AI reviews every line for bugs, security issues, and best practices
• 📝 Auto-Generate Descriptions - Never write PR descriptions again
• ⚡ Speed Up Reviews - Auto-approve safe changes, focus human review on what matters
• 🛡️ Security First - Built-in security analysis and safe auto-approval logic

License Notice: Modified version of PR-Agent by Codium AI, licensed under AGPL v3. Source code available here.

🎯 Core Features

Feature	Description	Benefits
🤖 Auto Review	Comprehensive code analysis with security scanning	Catch bugs, security issues, and code quality problems
📋 Auto Description	AI-generated PR titles, summaries, and labels	Save time, improve PR documentation
💡 Code Suggestions	Specific improvements and best practices	Learn better coding patterns, optimize performance
✅ Auto Approval	Safe approval of low-risk changes	Speed up workflow, focus reviews on complex changes
🎯 Smart Triggers	Flexible activation via PR events or manual commands	Control when and how the bot runs
📐 Repository Rules	Respects official Cursor rules files in your repository	Follows your project's coding standards automatically

🚀 Quick Setup

Basic Usage

name: AI PR Bot
on:
  pull_request:
    types: [opened, reopened, ready_for_review, synchronize]

jobs:
  ai-pr-bot:
    runs-on: ubuntu-latest
    steps:
      - uses: jacsamell/github-pr-bot@v1
        with:
          anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
          auto_review: true
          auto_describe: true
          enable_auto_approval: false

Advanced Configuration

name: AI PR Bot
on:
  pull_request:
    types: [opened, reopened, ready_for_review, synchronize]

jobs:
  ai-pr-bot:
    runs-on: ubuntu-latest
    steps:
      - uses: jacsamell/github-pr-bot@v1
        with:
          anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
          github_token: ${{ secrets.GITHUB_TOKEN }}
          auto_review: true
          auto_describe: true
          auto_improve: true
          enable_auto_approval: true
          model: 'anthropic/claude-sonnet-4-20250514'
          max_model_tokens: '1000000'
          require_trigger: false

🔑 Required Setup

1. Get Your Anthropic API Key

• Visit Anthropic Console
• Create an account and generate an API key
• Make sure you have Claude API access

2. Add Secret to GitHub

• Go to your repository → Settings → Secrets and variables → Actions
• Click "New repository secret"
• Name: ANTHROPIC_API_KEY
• Value: Your API key from step 1

Secret	Description	Required
ANTHROPIC_API_KEY	Your Claude API key from Anthropic	✅ Yes

📋 Inputs

Input	Description	Required	Default
anthropic_api_key	Anthropic API key for Claude AI	✅ Yes	-
github_token	GitHub token for API access	❌ No	${{ github.token }}
auto_review	Enable automatic code review	❌ No	true
auto_describe	Enable automatic PR description generation	❌ No	true
auto_improve	Enable code improvement suggestions	❌ No	true
enable_auto_approval	Enable automatic approval of safe changes	❌ No	false
model	AI model to use	❌ No	anthropic/claude-sonnet-4-20250514
max_model_tokens	Maximum tokens for AI model	❌ No	1000000
require_trigger	Require ##prbot trigger in PR description	❌ No	false

📤 Outputs

Output	Description
review_posted	Whether a review was posted
description_updated	Whether PR description was updated
improvements_posted	Whether improvement suggestions were posted
auto_approved	Whether PR was automatically approved

Optional Configuration File

Create .pr_bot.toml:

[config]
model = "anthropic/claude-sonnet-4-20250514"
enable_auto_approval = true
max_model_tokens = 1000000
use_cursor_rules = true  # Enable Cursor rules (default: true)

[github_action_config]
require_aidesc_trigger = true  # Requires ##prbot in PR description
auto_describe = true
auto_review = true
auto_improve = true

📐 Cursor Rules Support

The bot automatically detects and respects official Cursor rules in your repository, ensuring AI reviews follow your project's specific coding standards.

Supported Rules Files

The bot looks for these official Cursor rules files:

Current Format (Recommended):

• .cursor/rules/*.mdc - Modern Cursor project rules files

Legacy Format (Deprecated but supported):

• .cursorrules - Legacy Cursor rules file

Example Rules File

Create .cursor/rules/style.mdc in your repository:

---
description: Code style and formatting rules
alwaysApply: true
---

# Project Coding Rules

## Code Style
- Use 2 spaces for indentation
- Prefer const over let when possible
- Always use semicolons
- Use single quotes for strings

## Naming Conventions
- Use camelCase for variables and functions
- Use PascalCase for classes and components
- Use UPPER_SNAKE_CASE for constants

## Patterns to Avoid
- Don't use console.log in production code
- Avoid nested ternary operators
- Never commit commented-out code

## Preferred Patterns
- Use TypeScript strict mode
- Prefer async/await over promises.then()
- Use destructuring for object properties

The bot will automatically include these rules in its analysis, ensuring consistent code reviews that match your project's standards.

By default, up to 10% of the model context (capped at 100k tokens) is allocated to Cursor rules. On 1M-context models, this allows up to 100k tokens of rules to be included.

Usage

Automatic Mode

• Runs on all PRs (default)
• Or add ##prbot to PR description if require_aidesc_trigger = true

Manual Commands

• /review - Get code review
• /describe - Generate PR description
• /improve - Get code suggestions

🔒 Auto-Approval Safety

The bot uses intelligent criteria to safely auto-approve low-risk changes:

✅ Safe to Auto-Approve

• Documentation updates and README changes
• Test additions and improvements
• Minor refactoring and code cleanup
• Configuration file updates
• Dependency updates (non-breaking)

❌ Requires Human Review

• Critical business logic changes
• Security-related code modifications
• Database schema changes
• API contract modifications
• Complex algorithmic changes

Safety First: The bot errs on the side of caution - when in doubt, it requests human review.

🔒 Security & Privacy

• API Keys: Your Anthropic API key is securely handled through GitHub Secrets
• Code Privacy: Code is only sent to Anthropic's Claude API for analysis
• No Storage: No code or data is permanently stored by this action
• Permissions: Only requires standard GitHub token permissions for PR operations

🤝 Contributing

Contributions are welcome! This project is based on PR-Agent and maintains AGPL v3 licensing.

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

📄 License

This project is licensed under AGPL v3 - see the LICENSE file for details.

Based on PR-Agent by Codium AI.

🆘 Support

• 📖 Documentation
• 🐛 Report Issues
• 💬 Discussions

Development

# Install dependencies
pip install -r requirements.txt

# Set environment variables
export ANTHROPIC__KEY="your-api-key"
export GITHUB_TOKEN="your-github-token"

# Run on a specific PR
python -m pr_agent.cli --pr_url=https://github.com/owner/repo/pull/123 review