LaReview

What is LaReview?

LaReview is a local-first code review workbench that turns a pull request or unified diff into a structured review plan and high-signal feedback using AI. Instead of producing comment spam, it helps reviewers understand a change, plan what to check, and generate focused review threads anchored to specific lines.

It’s designed to fit alongside your existing AI coding agent: you provide a PR URL or diff, LaReview fetches the necessary data locally, generates tasks and a review plan, and then lets you run the review and push feedback back to GitHub/GitLab.

Key Features

• AI-powered planning from PRs or diffs: Provide a GitHub/GitLab PR URL or a unified diff; LaReview analyzes intent and builds a task tree to guide the review.
• Task-focused review view: Review tasks are grouped by flow and ordered by risk, with progress tracking and navigation support including a files heatmap.
• High-signal, rule-aware feedback: The agent proactively identifies bugs and anchors feedback threads to specific lines, aiming to avoid unstructured “comment dump” behavior.
• Custom review rules: Define standards such as “DB queries must have timeouts” or “API changes need a migration note” so the review is aligned with your team’s expectations.
• Local context for codebase search: Link local Git repositories so the agent can search your codebase without uploading intermediate data to a server.
• Git host sync for PRs: Submit review feedback directly to GitHub or GitLab PRs, including automatic summary generation.
• Visual diagrams of architectural/flow changes: Generate diagrams to help you understand the change at a structural level before diving into the code.
• Learning from rejected feedback: Mark suggestions as “ignored” during reviews; LaReview analyzes rejections to calibrate future review behavior and reduce low-value nitpicks.
• CLI workflow: Run reviews from the terminal by loading PRs by URL/identifier or piping a git diff into LaReview.

How to Use LaReview

1. Install LaReview (the site notes a free, open source option via Homebrew for macOS, plus downloadable binaries for macOS and Linux/WSL).
2. Start a review input: Provide either a GitHub/GitLab PR URL (shown as owner/repo#123) or a unified diff (e.g., via piping git diff).
3. Choose or use an AI coding agent: LaReview can generate a review plan using a selected agent (examples listed include Claude, Codex, Gemini, etc.).
4. Generate the plan and review: LaReview generates a task tree, then you execute the review with notes, task/status tracking, and an export to Markdown.
5. Push feedback back to your PR (optional): LaReview can submit review feedback directly to GitHub/GitLab with an auto-generated summary.

Use Cases

• PR review planning for complex changes: When a pull request touches multiple flows, LaReview can group review tasks by flow and order them by risk to structure the deep-dive.
• Reviewing a unified diff quickly from a terminal: Developers can pipe git diff into LaReview to generate a plan and feedback without leaving the command line.
• Enforcing team standards during reviews: Teams can encode concrete rules (e.g., database query timeouts or API migration notes) so the review checks for these items consistently.
• Understanding architectural/behavioral impact: Before reading every file, reviewers can use generated diagrams to map the change and focus attention on relevant areas.
• Calibrating review quality over time: During ongoing reviews, marking suggestions as “ignored” lets LaReview learn patterns from rejections to reduce repeated low-value feedback.

FAQ

• Is LaReview local-first? Yes. The workflow described on the site includes local fetching of PR data via GitHub/GitLab CLI (gh/glab) and local context linking to search a codebase.

• What inputs does LaReview accept? It supports a GitHub/GitLab PR URL (example format shown as owner/repo#123) and unified diffs (e.g., via piping git diff).

• Does LaReview work with my existing AI coding agent? The site states it works with your existing AI coding agent, and it lists supported agents that can be used for generation.

• Can feedback be posted back to GitHub or GitLab? Yes. LaReview can sync review feedback directly to GitHub/GitLab PRs and generate a summary.

• Does LaReview export outputs anywhere? The review flow mentions exporting to Markdown.

Alternatives

• General-purpose AI code review chat tools: These focus on generating review comments, but LaReview is positioned as a reviewer-first workbench with a structured plan, task view, and emphasis on anchoring feedback to specific lines.
• Manual checklist-based review processes: Teams can use static review templates and checklists; LaReview differs by generating a task tree and review plan from the PR/diff and supporting custom rules.
• Diff/PR analysis tools focused on static insights: Tools that summarize changes or highlight risky code help with understanding, while LaReview emphasizes interactive planning, rule-aware feedback threads, and optional diagram generation.