LaReview
A free, local-first desktop app that turns a pull request diff into a structured review plan with diagrams and rule checks — using the AI coding agent you already have installed, not a hosted service.
🔗 Visit LaReview
Description
Reviewing a large pull request usually means scrolling through a wall of changed files trying to reconstruct what the author was actually trying to do, then hoping you remember to check the same things you always check (did this migration get a rollback plan, does this API change need a doc update). LaReview is built to make that first pass faster and more consistent by turning the diff itself into a structured plan before you read a single line of code.
LaReview is an open-source, local-first desktop app for code review, built as a cross-platform tool in Rust. Instead of running its own AI model in the cloud, it connects to the coding agent you already have installed via the Agent Client Protocol (ACP) and fetches pull requests through your own GitHub CLI — meaning no servers, no uploading your code anywhere, and no separate AI subscription to pay for. It turns a diff into a structured review plan, generates architecture diagrams so you can see structural changes before reading line by line, and lets you define explicit rules ("DB queries must have timeouts," "API changes need a migration note") that it checks automatically, with feedback threads anchored to specific lines rather than comment spam. Review feedback can be submitted straight back to the GitHub or GitLab pull request. It's built and maintained by a solo developer and is fully free and open-source (MIT/Apache 2.0).
💬 Our review
The short version: if you already pay for an AI coding agent (Claude Code, Cursor, or similar) and want a better review workflow layered on top of it for free, LaReview is a low-risk download worth trying — it's not asking you to replace your agent or pay again.
Its positioning is genuinely different from hosted AI review bots like CodeRabbit or Qodo, which run their own model in the cloud and typically charge per seat or per repo: LaReview is local-first and reuses whatever agent you already have, so there's no double billing and no code leaving your machine unless your own agent already sends it somewhere. That's a real advantage for privacy-conscious teams or anyone already deep into a specific agent's subscription. The trade-off is that review quality depends entirely on the underlying agent you connect it to — LaReview is the workbench, not the brain — so a weak local model will produce a weak review plan. As a solo-maintained open-source project, it's also worth checking how actively it's kept up to date with ACP and GitHub CLI changes before making it a daily habit. Best fit: developers who already have Claude Code, Cursor or another agent installed and want a structured review layer for free. Weaker fit: teams wanting a fully managed, zero-setup hosted reviewer — that's still better served by CodeRabbit or GitHub's own AI review tooling.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Entièrement gratuit, sous licence MIT/Apache 2.0. Nécessite de posséder déjà un agent IA de codage (Claude Code, Cursor, etc.) pour fonctionner.
Pros
Gratuit et open-source (MIT/Apache 2.0), pas de double abonnement
Local-first — le code ne quitte pas la machine via l'outil lui-même
Réutilise l'agent IA déjà installé plutôt que d'en imposer un nouveau
Génère des diagrammes d'architecture et des règles personnalisées vérifiées automatiquement
Cons
La qualité de la review dépend entièrement de l'agent IA connecté — LaReview n'est que l'atelier, pas le cerveau
Maintenu par un développeur solo — vérifier l'activité du projet avant d'en dépendre au quotidien
Nécessite déjà un agent IA de codage installé, inutile seul
Pas de solution hébergée clé en main pour les équipes qui veulent zéro configuration