roborev

roborev reviews commits in the background for developers working with AI coding agents. It brings findings back into active coding sessions so bugs can be addressed before a pull request.

• Continuous review Each commit triggers a review, with severity and file locations attached to findings that stay open until a person or agent closes them.
• Agent choice Use Codex, Claude Code, Gemini, Copilot or Cursor, choose a different reviewer from the code writer, or use local models through Ollama or LiteLLM.
• Review and repair Agents can apply fixes and commit them, while branch refinement repeats review and repair in an isolated worktree.
• Review panels Combine several reviewers into one verdict, including security and design checks, or analyze existing code for duplication, complexity and dead code.
• Persistent findings Terminal and browser interfaces show review history, disagreements and usage analytics; verification checks findings against current code and removes those that no longer apply.

It's open source under MIT and runs on macOS, Linux and Windows. The daemon stores history in local SQLite; optional PostgreSQL sync shares history across a team. It doesn't require a hosted roborev service, but cloud-backed agents send requests to their providers. Anonymous daemon telemetry can be disabled.

A daemon on a machine you own can also review GitHub pull requests and GitLab merge requests and post combined feedback. Review and fix agents have shell access, and the project is intended for trusted codebases.