Pullrol

• 

  Pullrow

Code review is one of the most important parts of software development, yet it remains slow, manual, and inconsistent. Developers spend hours identifying bugs, security issues, and missing validations, often under time pressure. Pullrol is an autonomous AI agent that integrates directly into GitHub and reviews pull requests automatically.

It analyzes the full diff of a pull request and:

• Detects bugs and logic issues
• Identifies security vulnerabilities
• Generates precise inline comments
• Suggests real code fixes
• Creates test cases

• Automatically submits a full review (Approve or Request Changes) Pullrol transforms code review into a fast, intelligent, and automated workflow. Pullrol is built as a Chrome extension combined with a backend service and an AI agent powered by Airia:

• The Chrome extension extracts pull request data directly from GitHub

• The backend processes the diff and communicates with the AI agent

• The Airia agent analyzes the code and returns structured insights

• The system then uses the GitHub API to automatically submit reviews and comments

We also implemented fallback mechanisms to ensure reliability and a smooth user experience. One of the main challenges was GitHub’s restriction on UI automation. Direct interaction with buttons like "Submit review" is blocked for security reasons. To overcome this, we redesigned the architecture to use the GitHub API instead of DOM automation. Another challenge was handling inconsistent data from AI outputs, such as missing or invalid line numbers for inline comments. We solved this by adding validation layers and fallback logic. we're proud of:

• Successfully transforming a manual process into a fully automated workflow
• Building a real AI agent, not just a code assistant
• Integrating seamlessly with GitHub
• Delivering end-to-end functionality: analyze, comment, fix, test, and review
• Creating a system that feels instant and intuitive for developers We learned how to design AI systems that interact with real-world platforms like GitHub.

We also learned the importance of reliability, validation, and user experience when working with AI-generated outputs.

Most importantly, we learned that the future of development is not just AI assistance — but AI agents that take action.

We plan to extend Pullrol with:

• Multi-agent workflows for large-scale projects
• Deeper integration with enterprise tools (Slack, Teams, CI/CD)
• Advanced security and compliance analysis
• Team collaboration features
• Continuous learning from developer feedback

Our vision is to build a fully autonomous AI code review system for modern development teams

Built With

• airia-ai-agent-platform
• api
• chrome-extension-apis
• express.js
• github-rest-api
• groq
• javascript
• node.js
• python-(for-backend-logic)