🛡️ CodeSentinel

AI Code Review & Security Assistant — Powered by loopy-agent

✨ Features

🚀 Quick Start

# Install
pip install codesentin3l

# Or clone and install
git clone https://github.com/Dream-Pixels-Forge/codesentin3l
cd codesentinel
pip install -e .

📖 Usage

CLI Commands

# Security scan (no API key needed)
codesentin3l scan myfile.py

# AI-powered review
export OPENAI_API_KEY="sk-..."
codesentin3l review myfile.py

# Full scan + review
codesentin3l full myfile.py

# Scan entire repository
codesentin3l repo ./my-project

# Review a git diff
codesentin3l pr changes.diff

# Start MCP server
codesentin3l mcp

# Run demo
codesentin3l demo

CLI Options

--provider openai|anthropic    # LLM provider (default: openai)
--model gpt-4|claude-3-opus   # Model (default: gpt-4)
--output <dir>                 # Output directory for reports
--format html|markdown|json    # Report format

🐍 Python API

Quick Review

import asyncio
from src import CodeReviewer

async def main():
    reviewer = CodeReviewer(provider="openai", model="gpt-4")
    result = await reviewer.review_file("my_code.py")
    reviewer.print_report(result)
    await reviewer.close()

asyncio.run(main())

Security Scan

from src import SecurityScanner

scanner = SecurityScanner()
findings = scanner.scan_file("my_code.py")
scanner.print_findings(findings)

risk_score = scanner.get_risk_score(findings)

Repo Scanning

from src import RepoScanner

scanner = RepoScanner()
summary = scanner.scan_repo("./my-project")
scanner.print_summary(summary)

Report Generation

from src import ReportGenerator, ReportConfig, ReportFormat

config = ReportConfig(
    format=ReportFormat.HTML,
    output_dir="./reports",
)
generator = ReportGenerator(config)
content = generator.generate(review=result, security_findings=findings)
path = generator.save(content, "my-review")

🌐 MCP Server

Start the MCP server to expose CodeSentinel as tools:

codesentin3l mcp

Available Tools

Using with Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "codesentin3l": {
      "command": "python",
      "args": ["-m", "src.mcp_server"],
      "cwd": "/path/to/codesentinel"
    }
  }
}

🔗 GitHub Integration

Review a PR

from src import GitHubIntegration

async def main():
    gh = GitHubIntegration(token="ghp_...")
    review = await gh.review_pr("owner", "repo", pr_number=123)
    print(review.summary)
    print(f"Approved: {review.approved}")

asyncio.run(main())

Post Review Comment

review = await gh.review_pr("owner", "repo", 123, post_comment=True)

🔒 Security Patterns Detected

📊 Report Formats

🏗️ Architecture

┌─────────────────────────────────────────────────────────┐
│                    CodeSentinel                         │
├─────────────────────────────────────────────────────────┤
│                                                         │
│  ┌─────────────┐    ┌─────────────┐    ┌─────────────┐ │
│  │   Security  │    │     AI      │    │   GitHub    │ │
│  │   Scanner   │    │   Reviewer  │    │ Integration │ │
│  │  (patterns) │    │  (loopy)    │    │   (PRs)     │ │
│  └──────┬──────┘    └──────┬──────┘    └──────┬──────┘ │
│         │                  │                  │         │
│         └──────────────────┼──────────────────┘         │
│                            │                            │
│                    ┌───────▼───────┐                    │
│                    │    Reporter   │                    │
│                    │ HTML/JSON/MD  │                    │
│                    └───────────────┘                    │
│                                                         │
├─────────────────────────────────────────────────────────┤
│                   loopy-agent                           │
│  ┌──────────┬──────────┬──────────┬──────────┐         │
│  │   loop   │ gateway  │ verification│  cache  │         │
│  └──────────┴──────────┴──────────┴──────────┘         │
└─────────────────────────────────────────────────────────┘

🧪 Examples

See the examples/ directory:

• vulnerable_app.py — Code with intentional security issues
• clean_app.py — Secure coding best practices

📝 License

MIT

🤝 Contributing

Contributions welcome! See CONTRIBUTING.md.

Built with ❤️ using loopy-agent