LaReview

LaReview is a local-first, open-source code review workbench that transforms GitHub and GitLab pull requests or raw diffs into structured, flow-based review plans. Unlike bot-driven tools that flood PRs with comment spam, LaReview acts as a staff-engineer-level reviewer, identifying system-level hazards and grouping tasks by flow and risk. It works with your existing AI coding agents (Claude, Codex, Gemini, Mistral, Qwen, and more) and never sends your code to intermediate servers.

• AI-Powered Planning — Paste a PR URL or diff and LaReview generates a structured review checklist, identifying flows and hazards like a staff engineer.
• Task-Focused Review — Tasks are grouped by flow and ordered by risk, with a files heatmap to track progress and navigate changes efficiently.
• High-Signal Framework — The AI proactively identifies bugs and validates them against your custom rules, anchoring feedback threads to specific lines of code.
• Custom Rules — Define standards like "DB queries must have timeouts" or "API changes need a migration note" to enforce them automatically on every review.
• Local Context — Link local Git repositories so the agent can search your codebase without uploading any files, keeping data private.
• Git Host Sync — Submit review feedback directly to GitHub or GitLab PRs with automatic summary generation.
• Visual Diagrams — Auto-generate architectural diagrams to visualize flow changes before reading a single line of code.
• Learning Patterns — Mark suggestions as "ignored" to teach the AI your preferences; it analyzes rejections over time to reduce nitpicks and improve signal.
• CLI Support — Launch reviews from the terminal with lareview, pipe a git diff, or load a PR by URL directly from the command line.
• Zero Data Leaks — All data is fetched locally via the GitHub/GitLab CLI (gh/glab) with no intermediate servers involved.