Vercel Launches Open-Source AI Code Review Application
Guillermo Rauch, CEO of Vercel, announced OpenReview on March 5, 2026—an open-source, self-hosted application that automates pull request reviews using Claude AI integrated with Vercel's infrastructure. The tool runs in isolated sandbox environments and can analyze code, suggest improvements, and directly commit fixes to branches.
OpenReview Combines Code Analysis With Direct Fix Capabilities
Unlike traditional code review tools that only provide suggestions, OpenReview can directly fix formatting errors, lint issues, and simple bugs, then commit and push changes to PR branches. The workflow begins when developers mention @openreview in pull request comments, triggering the system to spin up an isolated sandbox, clone the repository, and deploy a Claude-powered agent to analyze changes.
Core capabilities include:
- Security issue identification
- Running project linters
- Explaining architectural patterns
- Committing formatting and lint fixes automatically
- Interactive feedback through emoji reactions (👍❤️ to approve, 👎😕 to skip)
Progressive Skill System Prevents AI Overwhelm
OpenReview uses extensible "skills"—specialized review modules stored in .agents/skills/—that load contextually to maintain focused analysis. This modular approach prevents the AI from attempting to analyze everything simultaneously, improving both performance and accuracy. The skill system allows teams to customize review priorities based on project needs.
Built on Vercel's Own Infrastructure Stack
The application leverages Vercel Workflow for durable execution, Vercel Sandbox for isolated code running, and the AI SDK for Claude integration. Rauch described the project as taking "everything I learned building Ultracite Cloud (closing soon) and turning it into a template that we can use internally to dogfood our primitives; and you can use."
The GitHub repository accumulated 341 stars as of March 2, 2026, while Rauch's announcement received 293 likes, 16 retweets, and 253 bookmarks on X. The project is currently in beta, with Vercel using it internally to test their own platform capabilities while making it available to the broader developer community.
Key Takeaways
- Vercel CEO Guillermo Rauch released OpenReview, an open-source AI code review bot that runs in isolated sandboxes and can directly commit fixes
- The system uses Claude AI with a modular "skills" architecture stored in
.agents/skills/to maintain focused, contextual analysis - OpenReview can identify security issues, run linters, and automatically fix formatting and simple bugs with developer approval via emoji reactions
- Built on Vercel Workflow, Vercel Sandbox, and the AI SDK, the tool serves as both an internal dogfooding exercise and public template
- The GitHub repository reached 341 stars as of March 2, 2026, with the project currently in beta