LaReview

LaReview provides a local‑first workbench for reviewing code changes. It ingests a unified diff or a pull‑request URL from GitHub or GitLab, then uses an attached AI coding agent to generate a structured review plan that highlights flows, hazards, and high‑risk areas. The plan is presented as a set of tasks, a heatmap of affected files, and optional visual diagrams that illustrate architectural changes before any line‑by‑line reading.

The tool is aimed at senior engineers who prefer depth over rapid comment dumping. It supports custom rule definitions—such as requiring timeouts on database queries or migration notes for API changes—and enforces them without producing spammy comments. Feedback is anchored to specific lines, and a learning component adjusts future suggestions based on which items are ignored or rejected. Reviews can be conducted in the browser or via a CLI that streams diffs directly from the terminal.

LaReview synchronizes review results back to the originating Git host, automatically summarizing feedback for the pull request. It keeps all code analysis local, linking to the developer’s repository to avoid uploading source material, and works with a range of AI agents, including Claude, Codex, Gemini, Kimi, Mistral, OpenCode, and Qwen.