">
[go: up one dir, main page]

Review AI Code Before Commit

GitHuman gives AI coding agents a clean handoff to human reviewers.
Review staged changes, collect feedback, and send it back to the agent.

cd /your/repo && npx githuman@latest ask "Please review these changes"

The Problem

AI coding agents write code. But the traditional PR workflow assumes humans are the authors. By the time AI-generated code reaches a pull request, you've already committed to the approach. Review happens too late.

The Solution

GitHuman moves the review checkpoint to where it belongs: the staging area. Before git commit, you get a proper review interface—not a wall of terminal diff output.

A GitHub-like interface for your staging area

GitHuman staged changes review interface

Features

Visual Diff Review

Review staged changes in a clean, syntax-highlighted interface. No more squinting at terminal output.

Agent-to-Human Handoff

Run githuman ask to open GitHuman, wait for a human review, and hand the resulting comments and todos back to the agent.

Inline Comments

Add comments to specific lines with optional code suggestions. Track issues before they become commits.

Review Workflow

Track status: in progress, approved, or changes requested. Make informed decisions about AI output.

Todo Tracking

Create tasks for follow-up work via CLI or web interface. Keep track of what needs attention.

Markdown Export

Export reviews with comments for documentation. Keep a record of your review decisions.

Local & Private

Everything runs on your machine. No data leaves your computer. No accounts required.

Agent Skills

Run npx skills add mcollina/githuman-skills to teach AI agents when and how to use GitHuman. Learn more at skills.sh.

Mobile Ready

Fully responsive interface. Review code on your phone or tablet when you're away from your desk.

Works beautifully on mobile

Review AI-generated code from anywhere. The interface adapts seamlessly to any screen size.

GitHuman mobile staged changes view GitHuman mobile review view

How It Works

1

AI agent makes changes

Claude, Copilot, Cursor, or any tool stages code changes

2

Run githuman ask

GitHuman starts or reconnects, prints the URL, and can open the browser automatically

3

Human reviews the changes

They inspect the diff, add comments, and create follow-up todos in the UI

4

Click Continue assistant

That explicit action tells GitHuman the human's turn is complete

5

The agent gets structured feedback back

githuman ask exits with the new todos, comments, and review status so the agent can continue