7 setups · GitHub Actions, overnight PRs, cross-review
AI code review agents: 7 pull request setups
Last updated · by Thomas, editor · How we verify
These aren't product rankings. They're setups people and teams run themselves: a bot that reviews every PR, a night shift that turns labeled bugs into draft PRs, and the simple rule that whoever wrote the code doesn't review it.
What they share is permissions: reviewers comment, humans merge, and the limits are enforced by GitHub settings, not just by the prompt.
Which one fits your repo
Where each reviewer runs, what starts it, and how much it may change.
@claude in pull requests (GitHub Action)
Runs on GitHub Actions, Anthropic API key. Starts on: PR or @claude mention.
⏱ Setup: 15 min✋ Limit tools so it comments, not pushesYour own PR review bot with Codex
Runs on GitHub Actions, OpenAI API key. Starts on: PR opened.
⏱ Setup: 15 min✋ contents: read; only the comment step writesOvernight test coverage PRs with Codex
Runs on Codex connected to GitHub. Starts on: Queued before you log off.
⏱ Setup: 20 min✋ Opens PRs; humans mergeNight shift: labeled GitHub bugs → reviewed draft PRs by morning
Runs on Grok Bot, GitHub, cloud coding agents, CI. Starts on: Nightly, on a label.
⏱ Setup: Half a day to set up✋ Draft PRs only; branch protection blocks mergesSplit a big task into a DAG of small subagents
Runs on Claude Code plugin, Node.js, git. Starts on: Per large task.
⏱ Setup: 20 min✋ Dispatch after you approve the graphThe writer never reviews its own lane
Runs on Agents from two or more vendors. Starts on: Every change.
⏱ Setup: 10 min✋ Nothing merges without a non-author passClaude Code builds, Codex reviews
Use one agent as the builder and the other as a second pair of eyes that hunts bugs and challenges architecture before you accept the result.
⏱ Setup: 5 min✋ The reviewer is read-only
Details come from each recipe's source. Setup time is shown only where the recipe states it.
Review bots on every PR
Official GitHub Actions you install in your repo.
@claude in pull requests (GitHub Action)
Anthropic's official action runs Claude Code on your own GitHub runner to review PRs, answer @claude questions and implement small changes.
Your own PR review bot with Codex
OpenAI's action runs Codex CLI inside a workflow with locked-down privileges; the README's example posts a code review comment on each new PR.
Agents that open PRs for you
Work happens overnight; you review in the morning.
Overnight test coverage PRs with Codex
OpenAI's internal guide lists seven Codex use cases; one engineer points it at low-coverage modules overnight and reviews runnable test PRs in the morning.
Night shift: labeled GitHub bugs → reviewed draft PRs by morning
A Chief bot and four specialists turn labeled GitHub bugs into reproduced, fixed, independently reviewed draft PRs overnight, with a morning report that separates done from attempted.
Split a big task into a DAG of small subagents
A Claude Code plugin that breaks work into atomic tasks with dependencies and gives each subagent only the context it needs, in parallel worktrees.
Review by a different agent
A second model that didn't write the code.
The writer never reviews its own lane
A fleet rule: whichever model wrote a piece of work never reviews it; a different vendor's agents review every lane, and repeated failure forces a redesign.
Claude Code builds, Codex reviews
Use one agent as the builder and the other as a second pair of eyes that hunts bugs and challenges architecture before you accept the result.
How to choose
- Trigger on mention on busy repos. The Claude action's failure list warns costs grow if it runs on every PR. (recipe)
- Treat PR text as untrusted. Prompt injection from PR or issue text is the first listed failure; keep tokens read-only.
- Enforce 'never merge' outside the prompt. The night shift protects the default branch and requires CI so the rule holds even if the agent ignores it. (recipe)
- Read the assertions. Overnight test PRs can lock in current bugs if the tests only assert today's behavior. (recipe)
Common questions
Should an AI agent be allowed to merge PRs?
None of these setups allow it. They comment or open draft PRs, and a person merges; the night-shift setup enforces that with branch protection.
Claude or Codex for PR review?
Both have official GitHub Actions, and both are listed here with their permission setups. Several recipes deliberately use a different model to review than the one that wrote the code.
Get the next money recipe. Free.
New recipes every week, plus the top 10 agent setups pack. One email, no spam.
Unsubscribe anytime. Privacy
More guides: Make money with AI agents · AI voice agent business · AI agent agency ideas · Claude Code use cases · Grok Bot use cases · Claude Code vs Codex · AI agents for small business · AI email agents · Autonomous agents that made money