AI wrote the PR.
You still have to review it.

The human review layer for AI-written pull requests.

Claude drafts the review from the real diff on your own plan: nothing posts until you click.

GitHub
  • Open source, MIT
  • Runs on your Claude plan
  • Mac and Linux, plus your phone

A reviewer's assistant, not a review bot.

Review bots

  • Writes straight into the PR, under its own account
  • Gets muted within weeks
  • Can approve, and nobody owns it
  • A separate bill, per seat or per review

ReviewStage

  • Drafts privately, for your eyes only
  • You post one review, as you
  • Approve is always your click
  • Runs on the Claude plan you already pay for

Four steps, and only one of them is the AI's.

  1. 1Requested

    A request lands in your queue and phone. Nothing runs until you click.

  2. 2Drafted

    Claude Code reads the real diff on your plan, at your chosen effort.

  3. 3Staged

    Start with a reading guide, then check each finding against the code. Keep what's worth your name, edit it, drop the rest.

  4. 4Posted

    One plain comment review, under your GitHub account. Approval is a separate click.

Everything a reviewer needs, nothing that speaks for you.

The QR code the desktop app shows for signing a phone in

Your phone, signed in by a QR code

Scan once. Get a push when a review is ready, swipe to keep, tap Post.

One command, a real app

A desktop window, a menu-bar icon, one-restart updates. No Docker.

It learns your team

Drop the same complaint three times and it drafts a rule for you to accept.

Findings you can check.

Each finding shows where to look, what to check, and a Checked, Likely, or Maybe label to guide your verification.

Two reviewers, independent runs

Separate runs per reviewer. Findings both raised are marked confirmed.

For a team

One server, everyone signed in as themselves, many repositories. Requests go to Slack, Discord or any webhook; Insights shows what the team keeps.

You decide what becomes a review

  • The AI draft stays private until you choose to publish.
  • You decide which findings are correct and worth saying.
  • The review you publish appears under your GitHub identity.
  • Approving a pull request requires a separate human action.
  • The drafting agent cannot write to GitHub.

Enforced, not promised.

A prompt-injected diff cannot post under your name: the agent cannot reach GitHub, and the run checks nothing landed.

  • No credentials

    Every GitHub token is removed before the agent starts.

    env -u GH_TOKEN -u GITHUB_TOKEN \
        -u GITHUB_PAT … claude -p
  • No tools that write

    A deny list the prompt cannot lift.

    --disallowedTools "Bash(gh:*)
      Bash(git push:*) Bash(curl:*) …"
  • Proof afterwards

    Reviews, comments and threads are counted before and after; any change fails the run.

    gh_before=$(pr_write_fingerprint)
    …
    [ "$gh_before" != "$gh_after" ] && fail
Read the security model

A private draft. Your decision to publish.

ReviewStageAuto-posting review botsReviewing alone
Who posts You, when you clickThe tool, into the PRYou
Whose name is on it YoursThe bot'sYours
Approves on its own NeverMost don't; some canNever
Learns what you drop Yes, as rules you acceptVariesOnly you do
Cost Your Claude planPer seat or per reviewYour time
Where your data lives Your machine or your serverThe vendor's cloudYour machine

The diff goes to Anthropic via Claude Code on your account; what you post goes to GitHub. Full list: What leaves my machine? below.

Start on your laptop. Bring the team when it's worth it.

Just me

  1. 1Continue with GitHub
  2. 2Connect Claude
  3. 3Pick repositories

Scan the QR code in the app menu to add your phone. Needs Node 20+, Git and Claude Code. Not ready to install? Try a stored review live.

My team

docker compose up -d
  • One server for everyone
  • Each reviewer signs in as themselves
  • Each run on the reviewer's own Claude plan

Same gate at every size: nothing reaches GitHub until a person clicks.

Questions people ask first.

What does it cost?

Your existing Claude plan. No API key, no per-review bill, and no server bill for the desktop app.

What leaves my machine?

The diff and the files it touches go to Anthropic through Claude Code on your account; what you post goes to GitHub with your token. Beyond that: GitHub for sign-in and reading the PR, npm for update checks, Apple's and Google's push services if you turn phone notifications on (never the finding text), and Cloudflare's tunnel while phone access is on. Usage counts stay on your machine unless you opt in and a reporting endpoint is configured. No endpoint is configured by default.

Can it approve on its own?

No, not as an option, not behind a flag. Approve is your click.

Does it work with my repositories?

GitHub only: any repository you can review, public or private, as many as you like.

Does it run on Windows?

The Docker install runs under WSL2 today; a native app is on the roadmap.

How is this different from Claude Code Review, CodeRabbit or Copilot review?

Most AI review products deliver AI-written feedback directly into the pull request. ReviewStage keeps the draft private first. The reviewer decides what is correct, worth saying, and worthy of their name.

All questions

You still have to review it.

Let Claude write the first draft.

Star on GitHub