How to Set Up Claude Code Review

Prerequisites — What You Need Before Starting

Before enabling Claude Code Review, you need to prepare the following:

Step 1 — Enable Code Review from Admin Settings

The first thing to do is enable the feature from Anthropic's Admin page:

1. Go to claude.ai/admin-settings/claude-code
2. Scroll down to the "Code Review" section
3. Click "Setup" to begin the GitHub App installation process

The system will redirect you to GitHub to install the Claude GitHub App.

Step 2 — Install the Claude GitHub App

Claude Code Review works through a GitHub App called "Claude" — it must be installed in your Organization first:

1. Go to github.com/apps/claude
2. Click "Install"
3. Select the desired Organization
4. Select Repositories — you can choose "All repositories" or select specific ones
5. Review the Permissions requested by the App:

6. Click "Install & Authorize"

Step 3 — Choose Trigger Mode

After installing the GitHub App, choose when Claude should Review:

Recommended: Start with "PR Creation Only" — then switch to "After Every Push" when you see the Reviews are truly useful.

Step 4 — Create CLAUDE.md for Review Rules

The standout feature of Claude Code Review is the ability to define Review rules using natural language via the CLAUDE.md file — no YAML or Regex required.

What is CLAUDE.md?

CLAUDE.md is a file placed at the Repository root to tell Claude:

• What Coding Standards this project uses
• What Architecture Decisions must be maintained
• What specific rules must always be checked (e.g., "never commit secrets")

# Project: E-Commerce Backend

## Tech Stack
- Python 3.12, FastAPI, PostgreSQL
- ORM: SQLAlchemy 2.0 (async)
- Testing: pytest + httpx

## Coding Standards
- Use Type Hints for every function
- Do not use print() in production code — use the logging module
- Every API Endpoint must have a Pydantic schema
- Database queries must always use parameterized queries

## Security Rules
- Do not hardcode API keys, passwords, tokens
- Input validation for every endpoint
- Rate limiting for public endpoints

## Architecture
- Clean Architecture: Controller → Service → Repository
- Do not import Repository layer directly from Controller

REVIEW.md — Review-Specific Rules

In addition to CLAUDE.md, you can create a REVIEW.md file (or .claude/review.md) to define rules used specifically during Code Review:

# Review Rules

## Always Check
- SQL Injection vulnerabilities
- Functions longer than 100 lines
- Missing error handling in async operations
- N+1 query patterns

## Ignore
- Formatting/style issues (handled by black + isort)
- Type annotation completeness (handled by mypy)
- Import ordering (handled by isort)

## Severity
- Security issues → always flag as Critical
- Missing tests for new endpoints → flag as High

.claude/rules/ — Modular Rules

For large projects, you can create Rules as separate .md files in .claude/rules/ — Claude loads all files automatically, equivalent to CLAUDE.md.

Step 5 — Test Your First Review

Once setup is complete, test by opening a PR:

1. Create a new Branch from existing code
2. Modify some code (try introducing a minor issue, e.g., remove error handling)
3. Push the Branch and open a Pull Request
4. Wait ~20 minutes — you'll see Check Run "Claude Code Review" appear
5. When the Review is complete, you'll see:

Step 6 — Read and Respond to Review Results

When Claude Code Review finishes, Comments will appear on the PR — here's how to handle them:

Reading Review Comments

• Summary Comment — read the overview first to see how many issues and at what severity levels
• Inline Comment — click to view each Finding on the problematic code lines
• Suggested Fix — some Findings include corrected code to review

Responding

Step 7 — Use CLI Mode for Local Review

Besides automatic Reviews via the GitHub App, you can also run Reviews directly from the Claude Code CLI:

# Review current PR (output in Terminal)
/code-review

# Review PR and Post Comment to GitHub
/code-review --comment

# Review a specific PR by number
/code-review --pr 123

Advantages of CLI Mode:

• Preview Reviews before pushing — no need to wait for a PR
• Choose to review specific PRs only — no need to review every PR
• Use with --comment to post results to GitHub immediately

Best Practices — Getting the Most from Claude Code Review

GitHub Actions — Advanced Workflow Configuration

For teams wanting more control, you can use GitHub Actions instead of the GitHub App:

name: Claude Code Review

on:
  pull_request:
    types: [opened, synchronize]

jobs:
  review:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
        with:
          fetch-depth: 0

      - name: Run Claude Code Review
        uses: anthropics/claude-code-review@v1
        with:
          anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
          github_token: ${{ secrets.GITHUB_TOKEN }}
          confidence_threshold: 80

Setting up the API Key:

1. Go to Repository → Settings → Secrets and variables → Actions
2. Click "New repository secret"
3. Name it ANTHROPIC_API_KEY
4. Paste the API Key from console.anthropic.com

Confidence Threshold — Adjusting Strictness

Confidence Threshold is a 0-100 number that determines how confident Claude must be before reporting an Issue:

Real-World Statistics from Anthropic's Usage

Anthropic uses Claude Code Review on every internal PR — results:

These numbers show that Claude Code Review significantly improves Review quality — especially for large PRs that people often don't have time to review thoroughly.

Claude Code Review in the Thai Enterprise Context

For Thai organizations with software development teams, Claude Code Review can help in many scenarios:

• Small teams without Senior Reviewers — use Claude as a "second Senior Developer" to review code
• ERP projects with large codebases — ERP systems often have hundreds of Modules, too many for human review alone
• Compliance with security standards — define in CLAUDE.md what to check, e.g., SQL Injection, 2FA
• Outsourced or AI-written code — verify code quality from external sources before Merge

Summary — Claude Code Review Setup Checklist

Claude Code Review doesn't replace Human Reviewers — it serves as a "second pair of eyes" that never gets tired, never forgets, and never skims through code — all the team needs to do is "tell it what to look for" via CLAUDE.md.

— Saeree ERP Team

Continue Reading — EP 1 and EP 3

• EP 1: What is Claude Code Review? — Multi-Agent Automated Code Review System
• EP 3: Claude Code Review vs Competitors + Price and Value

References

• Claude Code Official Docs — Code Review
• Claude Code Official Docs — GitHub Actions
• TechCrunch — Anthropic launches code review tool
• The New Stack — Anthropic launches multi-agent code review

If your organization is looking for an ERP system developed with high standards and a focus on software quality, you canschedule a demo or contact our advisory teamfor further discussion