How to Create an Automated Code Review System

An automated code review system with AI quality analysis catches bugs, style issues, and security problems before code ships. I run AI reviews on every commit in my projects. It is not a replacement for human code review on critical paths, but it handles the 90% of reviews that are about consistency and basic quality.

What You Need Before Starting

• Python 3.8+
• Anthropic API key
• Git installed
• A project repository

Step 1: Get the Diff to Review

import subprocess

def get_git_diff(base_branch="main"):
    result = subprocess.run(
        ["git", "diff", base_branch, "--unified=5", "--diff-filter=ACMR"],
        capture_output=True, text=True
    )
    return result.stdout

def get_staged_diff():
    result = subprocess.run(
        ["git", "diff", "--cached", "--unified=5"],
        capture_output=True, text=True
    )
    return result.stdout

Step 2: Build the Review Engine

import anthropic
from dotenv import load_dotenv

load_dotenv()

REVIEW_PROMPT = """Review this code diff. Focus on:
1. Bugs or logic errors
2. Security vulnerabilities (SQL injection, hardcoded secrets, etc.)
3. Performance issues
4. Missing error handling
5. Code style inconsistencies

For each issue found, respond with:
- File and line number
- Severity (critical/warning/suggestion)
- What the problem is
- How to fix it

If the code looks good, say so briefly. Do not invent problems that do not exist."""

def review_diff(diff_text, context=""):
    client = anthropic.Anthropic()
    
    message = f"Code context: {context}\n\n" if context else ""
    message += f"Review this diff:\n\n```diff\n{diff_text}\n```"
    
    response = client.messages.create(
        model="claude-sonnet-4-20250514",
        max_tokens=2048,
        system=REVIEW_PROMPT,
        messages=[{"role": "user", "content": message}]
    )
    
    return response.content[0].text

Step 3: Parse Review Results

import re

def parse_review(review_text):
    issues = []
    current_issue = None
    
    for line in review_text.split("\n"):
        severity_match = re.search(r"\**(critical|warning|suggestion)\**", line.lower())
        if severity_match:
            if current_issue:
                issues.append(current_issue)
            current_issue = {
                "severity": severity_match.group(1),
                "text": line.strip()
            }
        elif current_issue:
            current_issue["text"] += "\n" + line.strip()
    
    if current_issue:
        issues.append(current_issue)
    
    return issues

Step 4: Build the CLI

import argparse

def main():
    parser = argparse.ArgumentParser(description="AI Code Review")
    parser.add_argument("--staged", action="store_true", help="Review staged changes")
    parser.add_argument("--branch", default="main", help="Base branch for diff")
    parser.add_argument("--context", default="", help="Project context for better reviews")
    args = parser.parse_args()
    
    if args.staged:
        diff = get_staged_diff()
    else:
        diff = get_git_diff(args.branch)
    
    if not diff.strip():
        print("No changes to review.")
        return
    
    print(f"Reviewing {len(diff.splitlines())} lines of changes...\n")
    review = review_diff(diff, args.context)
    print(review)
    
    issues = parse_review(review)
    critical = [i for i in issues if i["severity"] == "critical"]
    if critical:
        print(f"\n{len(critical)} critical issues found. Fix before merging.")
        exit(1)

if __name__ == "__main__":
    main()

Step 5: Add as a Git Hook

Create .git/hooks/pre-push:

#!/bin/bash
echo "Running AI code review..."
python3 /path/to/code_review.py --branch origin/main
if [ $? -ne 0 ]; then
    echo "Critical issues found. Push blocked."
    exit 1
fi

chmod +x .git/hooks/pre-push

What to Build Next

Add a review history database. Track which files get the most issues over time. That tells you where to focus refactoring effort.

Related Reading

• The Testing Pyramid for AI Operations - quality layers for AI-powered development
• Claude Code: The AI Coding Tool That Changes Operations - AI in the coding workflow
• Building Your First Automation: A Complete Guide - automation fundamentals