Vercel Launches OpenReview: Open-Source AI-Powered Code Review Bot Powered by Claude
Key Takeaways
- ▸OpenReview enables on-demand AI code reviews triggered by mentioning @openreview in PR comments, powered by Claude Sonnet 4.6
- ▸The bot can execute code in sandboxed environments, run linters and tests, and automatically commit fixes for formatting and lint errors
- ▸Built on Vercel's infrastructure, the tool uses a progressive skill system and supports custom extensions for tailored code review workflows
Summary
Vercel has released OpenReview, an open-source, self-hosted AI code review bot that integrates with GitHub to provide on-demand automated code reviews. Powered by Anthropic's Claude Sonnet 4.6 model, the tool allows developers to trigger reviews by mentioning @openreview in pull request comments. The bot can analyze code diffs, run linters and tests in a sandboxed environment, post inline suggestions with one-click fixes, and even commit formatting and lint fixes directly to branches.
Currently in beta, OpenReview runs on Vercel's infrastructure, leveraging Vercel Workflows for durable, resumable execution and Vercel Sandbox for isolated code execution with full repository access. The platform features a progressive skill system that loads specialized instructions only when relevant, keeping reviews focused and efficient. Developers can customize behavior through simple Next.js route handlers and extend functionality with custom skills defined in .agents/skills/ directories.
The tool was built internally by Vercel as a way to test their technologies together and is now available for public deployment. Setup requires creating a GitHub App with appropriate permissions, configuring environment variables, and installing the app on target repositories. Developers can approve or skip suggestions using GitHub reactions, making the review workflow seamless and integrated into existing development processes.
- The open-source project is currently in beta and designed for self-hosting, allowing teams to integrate AI-assisted code review into their GitHub workflows
Editorial Opinion
OpenReview represents a practical application of Claude in developer workflows, addressing a real pain point in code review processes. The ability to run actual linters and tests in sandboxed environments gives it a significant advantage over simpler LLM-based review tools that can only analyze text. While the beta status suggests rough edges remain, the progressive skill system and extensibility features show thoughtful architecture for long-term adoption.
