hihi,

LaReview is a dev-first code review workbench for complex changes.

You give it a PR (GitHub/GitLab) or a diff, and it builds a structured review plan grouped by flows (auth, API, billing) and ordered by risk. The goal is to make big reviews feel like a plan you can actually follow, not an endless scroll.

It runs locally and is designed to work with your existing AI coding agent (bring your own agent). No bot comment spam. You decide what feedback gets posted back to the PR.

Highlights:

  - AI review planning: flow-based tasks + risk ordering
  - Task-focused diffs: isolate only the hunks relevant to one concern
  - Custom rules: enforce standards like “DB queries must have timeouts”
  - Optional diagrams to understand flows before reading code (requires D2)
  - GitHub/GitLab sync to submit selected feedback + generate a summary
  - Export summary to Markdown

CLI:

  lareview
  lareview pr owner/repo#123
  git diff | lareview
  lareview --agent claude

Install:

  brew install --cask puemos/tap/lareview

Repo:

  https://github.com/puemos/lareview

I would love feedback on the workflow, what you would want it to catch, and what would make you trust it in real reviews.