code-review

PackAI & models

code-review is a plugin that automates code review for pull requests. It runs several specialized review agents over each change and uses confidence-based scoring to filter out low-confidence findings, so the feedback a person sees is limited to issues the reviewers are actually confident about. It is aimed at code review for Claude plugin repositories.

Unavailable. Delivery for this kind is on the roadmap — not serving yet.

Have a repository with pull requests that need automated review.

What your AI can do with it

  • Automates code review of pull requests
  • Runs multiple specialized review agents on each change
  • Scores findings by confidence to filter false positives
  • Surfaces only high-confidence review findings
  • Targets review of Claude plugin repositories

Getting started

  1. Have a repository with pull requests that need automated review.
  2. Install the code-review plugin in your agent setup.
  3. Configure it against the repository or repos it should review.
  4. Open or update a pull request and let the plugin run its review agents.
  5. Check the reported findings, which are filtered by confidence score.

Signals

GitHub stars
148k
Forks
25k
Last commit
Sep 2026

Questions

What does code-review do?
It automates code review for pull requests by running multiple specialized review agents and using confidence-based scoring to filter out low-confidence findings.
How does it reduce false positives?
Each finding is assigned a confidence score, and findings below the confidence threshold are filtered out so only higher-confidence issues are reported.
Is it limited to Claude plugin repos?
It is built to automate code review for Claude plugin repositories, checking pull requests with several focused reviewers.
Does it review code itself or use agents?
It uses multiple specialized agents, each focused on reviewing a particular aspect of the pull request.
Advanced
Item type
plugin
Key
anthropics-claude-code-code-review
Source
github.com/anthropics/claude-code