Show HN: adamsreview – better multi-agent PR reviews for Claude Code
Claude Code plugin catches bugs that /review and CodeRabbit miss. One dev built it in parallel sub-agents. Now shipping.

Why it matters
A developer built a Claude Code plugin that outperforms existing AI code review tools (including Claude's own /review and /ultrareview) by orchestrating multi-stage, parallel agent reviews with persistent state management. It demonstrates how layering agent patterns on top of existing models can create production-grade tooling that competitors haven't matched.
The key facts
15 to knowadamsreview catches more real bugs than Claude built-in /review, /ultrareview, CodeRabbit, Greptile, and Codex built-in review
Six Claude Code slash commands: review, codex-review, add, promote, walkthrough, fix
Uses parallel sub-agents with persistent JSON state stored on disk
Includes ensemble review via Codex CLI and PR bot comments
Walkthrough command uses Claude's AskUserQuestion feature for human-in-the-loop validation
Fix command dispatches per-fix-group agents and re-reviews with Opus before committing
Runs on Claude Code Max plan subscription (unlike /ultrareview, which uses Extra Usage pool)
Published on Show HN (early-stage builder community)
adamsreview: Claude Code plugin with 6 slash commands (review, codex-review, add, promote, walkthrough, fix)
Uses parallel sub-agents, persistent JSON state, optional ensemble review
Claims to catch more real bugs with fewer false positives than Claude's built-in /review, /ultrareview, CodeRabbit, Greptile, Codex review
Runs on standard Claude Code Max subscription (not charged against Extra Usage pool like /ultrareview)
Employs multi-stage review architecture with context clearing between stages, human validation gates, regression detection
Published on Hacker News (Show HN) with early-stage adoption signals (8 points, 1 comment as of publish)
GitHub repo: adamjgmiller/adamsreview
Go to the source
Hacker Newsgithub.com
Publisher excerpt: I built adamsreview, a Claude Code plugin that runs deeper, multi-stage PR reviews using parallel sub-agents, validation passes, persistent JSON state, and optional ensemble review via Codex CLI and PR bot comments. On my own PRs, it has been catching dramatically more real bugs than Claude’s…