A Claude Code plugin that peer-reviews your code with AI agents. agentic-peer-review spawns an AI code reviewer (Claude, Codex, or Gemini), implements the recommended fixes, and repeats until the reviewer is satisfied with the codebase.
Open Claude Code and paste the following:
/plugin marketplace add gideonshaked/agentic-peer-review@v1.0.0
/plugin install agentic-peer-review@agentic-peer-review
Then run peer review on the current workspace:
/peer-review
From the terminal:
claude plugin marketplace add gideonshaked/agentic-peer-review@v1.0.0
claude plugin install agentic-peer-review@agentic-peer-reviewOr from inside Claude Code:
/plugin marketplace add gideonshaked/agentic-peer-review@v1.0.0
/plugin install agentic-peer-review@agentic-peer-review
From the terminal:
claude plugin marketplace add gideonshaked/agentic-peer-review
claude plugin install agentic-peer-review@agentic-peer-reviewOr from inside Claude Code:
/plugin marketplace add gideonshaked/agentic-peer-review
/plugin install agentic-peer-review@agentic-peer-review
From the terminal:
claude plugin marketplace update agentic-peer-review
claude plugin update agentic-peer-review@agentic-peer-reviewOr from inside Claude Code:
/plugin marketplace update agentic-peer-review
Then open /plugin, go to the Installed tab, and update the plugin from there.
- Claude Code
- uv
- Codex CLI (only required for using OpenAI Codex as a reviewer agent)
- Gemini CLI (only required for using Google Gemini as a reviewer agent)
To see all available options, in Claude Code run:
/peer-review -h
usage: peer-review [-h] [--agent {claude,codex,gemini}]
[--max-rounds MAX_ROUNDS] [--focus FOCUS]
[--timeout TIMEOUT] [--worktree] [--log LOG] [--only ONLY]
[instructions]
Iterative AI peer review that finds and fixes issues in your codebase
positional arguments:
instructions Optional review instructions
options:
-h, --help show this help message and exit
--agent {claude,codex,gemini}
AI agent to use for review (default: claude)
--max-rounds MAX_ROUNDS
Maximum review-fix cycles (default: 5). Stops early if
no issues found.
--focus FOCUS Narrow review scope to a specific file or directory
path
--timeout TIMEOUT Timeout in seconds for each review agent invocation
(default: 300)
--worktree Run fixes in a git worktree; show diff at end and ask
to merge or discard
--log LOG Write findings and fix/skip decisions to the specified
file
--only ONLY Comma-separated list of checks to run (default: all).
Available: architecture, bugs, dead-code, performance,
security, tech-debt
| Flag | Description |
|---|---|
-h, --help |
Show help message and exit |
--agent |
Review agent: claude (default), codex, or gemini |
--max-rounds |
Maximum review-that finds and fixes issues (default: 5). Stops early if no issues found. |
--focus <file or dir> |
Narrow the review to a specific file or directory |
--only <check,check,...> |
Only run specific checks. See checks below. |
--timeout |
Timeout per agent invocation in seconds (default: 300) |
--worktree |
Run all fixes in an isolated git worktree. Shows diff at end and asks to merge. |
--log |
Write a structured review log to the specified file |
instructions |
Give the reviewer agent specific instructions on what to focus on, what to skip, etc. |
By default all checks run. Use --only to select a subset:
| Check | What it looks for |
|---|---|
bugs |
Logic errors, off-by-one, null access, race conditions, unhandled edge cases |
security |
Injection, auth flaws, secrets in code, insecure defaults, missing validation |
dead-code |
Unreachable code, unused imports/variables/functions, obsolete config |
tech-debt |
Hardcoded values, copy-paste logic, missing error handling, TODO markers |
architecture |
Coupling, layering violations, circular dependencies, misplaced responsibilities |
performance |
Unnecessary allocations, N+1 queries, blocking calls, missing caching |