claude-code-review-council

This is a plugin for Claude Code, Anthropic's AI coding assistant, that adds a more thorough code review command. Instead of asking one AI model to look at your changes, it runs six reviewers at the same time and then combines their findings into a single report.

The six reviewers are: OpenAI's Codex CLI (using GPT-5.5 at its highest reasoning setting), Google's Gemini CLI (using Gemini 3.1 Pro), and four Claude specialist subagents each focused on a single concern: security, performance, logic, and regression risk. All six run in parallel, so the total wait time is roughly as long as the slowest reviewer, not the sum of all six. Each reviewer receives the same diff and the same context about what the change is supposed to do, which helps catch cases where the implementation does not match the stated intent.

After all six reports come in, a synthesis step combines them. It removes duplicate findings, checks each reported file and line number against the actual code to drop hallucinated citations, re-rates severity if a finding was over or under-stated, and flags places where reviewers disagreed. The result is a unified report organized by priority level. No code changes are made until the developer approves them.

The README explains why six reviewers help: different model families have predictable blind spots, and when models from different families agree on a finding, confidence in that finding is higher than when only one flagged it. If Codex or Gemini is unavailable due to quota limits or a network error, the four Claude specialists continue and complete the review on their own.

Installing requires the Codex and Gemini command-line tools to be set up separately. The plugin itself installs via the Claude Code plugin marketplace or by copying a folder. The README suggests using this for security-sensitive code, database migrations, and changes that would normally warrant a senior engineer's review, rather than for minor edits. The project is MIT-licensed.