LaReview

LaReview is a local-first code review workbench designed for senior engineers who prioritize depth and precision in their review process. Unlike conventional AI tools that generate comment spam or act as passive bots, LaReview transforms raw diffs into structured review plans, interactive diagrams, and actionable insights. The core philosophy is "Merge confidence, not just merge speed," emphasizing understanding and reliability over rapid but shallow approvals. Users begin by loading a diff or a pull request into the workbench. LaReview then analyzes the change and presents a high-level review plan, breaking down the modifications into logical sections, dependencies, and potential impact areas. This plan can be visually represented as flow-based diagrams, allowing reviewers to trace data changes, control flow alterations, or architectural shifts without scrolling through hundreds of lines of code. The tool supports multiple AI coding agents—including Claude Codex, Gemini, Kimi, Mistral, OpenCode, and Qwen—but integrates them in a controlled, reviewer-first manner. Instead of letting agents post automated comments directly, LaReview lets the engineer curate and organize feedback from these agents into a coherent, high-signal review. This prevents "comment dump" scenarios where irrelevant or repetitive comments clutter the conversation. The workbench is built with a "Plan First" approach: before writing any review comment, the engineer studies the structured plan, identifies the key areas needing attention, and then crafts feedback that is specific and meaningful. This results in fewer but more valuable comments. LaReview also emphasizes data privacy and security: it operates locally, meaning code changes never leave the user's machine. There are no cloud data leaks, no t