🚀 Developed using Cursor and vibe coding 🚀

Special thanks to:

• code2prompt - For code structure analysis
• cloc - For counting lines of code

AI Code Review Agent

An AI-powered code review agent that analyzes your codebase using large language model (LLM) for in-depth analysis and recommendations.

Features

• Project Structure Analysis - Uses code2prompt to generate detailed file structure of the project
• Technology Detection - Automatically identifies project type, languages, frameworks, and dependencies
• Intelligent Key File Detection - Identifies the most important files for in-depth review
• Context-Aware Analysis - Intelligently gathers related code when more context is needed
• Comprehensive Code Review - Analyzes code for:
  • Code quality issues
  • Architectural concerns
  • Security vulnerabilities
  • Performance implications
  • Best practices compliance
  • Commented code
• Project Size Evaluation - Uses cloc to count lines of code and categorize project size
• Detailed, Actionable Reports - Generates comprehensive reports with:
  • Project overview and structure
  • Prioritized issues with severity levels
  • Project strengths and positive practices
  • Specific, actionable recommendations
  • Dependency analysis
• Written in TypeScript for improved code quality and maintainability

Prerequisites

Choose your preferred installation method below, each with its specific prerequisites.

🐳 Usage with Docker (Recommended)

Prerequisites for Docker

• Docker

The easiest way to run the code review agent is with Docker, which includes all required dependencies:

1. Clone this repository

2. Copy .env.example to .env and add your Anthropic API key:

cp .env.example .env

Then edit the .env file to add your Anthropic or Google Gemini API key.

3. Build the Docker image:

docker build \
   --pull \
   --rm \
   -f "Dockerfile" \
   -t project-review-agent:latest \
   .

4. Run the container, mounting your project directory and the desired directory for the output:

docker run \
   -v "<PATH_TO_YOUR_PROJECT_DIRECTORY>:/project" \
   -v "<PATH_TO_THE_DESIRED_DIRECTORY_FOR_THE_OUTPUT>:/app/output" \
   project-review-agent:latest

💻 Alternative: Local Installation

Prerequisites for Local Installation

• Node.js (v22)
• Code2Prompt - v3.0.2
• cloc (Count Lines of Code)

If you prefer to run the tool without Docker:

1. Install all prerequisites listed above
2. Clone this repository
3. Install dependencies:

npm install

4. Copy .env.example to .env and add your Anthropic API key:

cp .env.example .env

Then edit the .env file to add your Anthropic API key.

5. Build the TypeScript project:

npm run build

6. Run the code review agent:

npm start -- [path-to-project]

Note: You must provide a path to the project directory to analyze.

For a quick test with the included test project:

npm run test

Output

The reports will be generated in the output directory, in a timestamped subfolder containing:

• project_review.md - A comprehensive analysis with all findings

How It Works

1. Project Structure Analysis - The agent first analyzes your project structure:

   • Uses code2prompt to generate a detailed file structure
   • Creates a comprehensive map of your codebase

2. Technology Detection - Based on the file structure:

   • Identifies project type (web application, mobile app, CLI tool, etc.)
   • Detects programming languages used
   • Determines frameworks and key dependencies

3. Key File Detection - The agent intelligently:

   • Identifies the most important files for review
   • Prioritizes files based on their purpose and potential impact
   • Creates a dependency map to understand relationships

4. Code Review - The agent then:

   • Uses Code2Prompt to prepare each file for analysis
   • Leverages the LLM for in-depth code review
   • Analyzes code quality, security, and architectural concerns
   • Examines related files when more context is needed

5. Project Size Evaluation - The agent:

   • Uses cloc to count lines of code
   • Categorizes the project size (small, medium, large)

6. Report Generation - Finally, it generates a comprehensive report with:

   • Project overview and technical details
   • Prioritized list of issues with severity ratings
   • Project strengths and positive practices
   • Specific, actionable recommendations
   • Dependency analysis and project statistics

Report Structure

The generated report includes:

1. Introduction - Overview of the project type, purpose, and technologies
2. Issues - Detailed analysis of problems found with severity levels
3. Strengths - Positive aspects and well-implemented practices
4. Miscellaneous - Project size, dependencies, and other metrics
5. Recommendations - Actionable suggestions for improvement

Environment Variables

The following environment variables can be set in your .env file:

• GOOGLE_API_KEY - Your Google API key (required)
• LangSmith configuration (optional)
  • LANGCHAIN_TRACING_V2
  • LANGCHAIN_ENDPOINT
  • LANGCHAIN_API_KEY
  • LANGCHAIN_PROJECT

License

MIT License. See LICENSE file for details.