🧙‍♂️ MageTool

The Open-Source Media Processing Suite & AI Agent Developer Toolkit.
Client-Side FFmpeg • Graphic Canvas Editor • Codebase Intelligence • Model Context Protocol (MCP)

Features • Web Platform • AI Agent CLI & MCP • Architecture • Quickstart • Contributing

🌟 Overview

MageTool is an open-source, full-stack developer and media workspace designed for modern creators, software engineers, and AI coding agents.

The project bridges browser-based, high-performance media manipulation with automated developer workflows:

1. Web Media & Document Platform: In-browser client-side media utilities powered by FFmpeg WebAssembly, Fabric.js canvas editor, image/video manipulation, and Cloudflare Edge microservices.
2. AI Agent Model Context Protocol (MCP) & CLI: A zero-bloat TypeScript developer CLI and stdio JSON-RPC MCP server enabling AI coding agents (OpenAI Codex, Claude, Cursor, Windsurf, Antigravity) to inspect codebases, detect security leaks, and perform automated PR reviews.

✨ Features

🎬 Web Media Platform (/frontend)

• Video Studio: Trimmer, video-to-GIF converter, rotate, frame extractor, speed adjuster, audio stripper, audio merger.
• Audio Engineering: Audio compressor, format converter, music overlay, metadata inspector.
• Image Processing: Passport photo generator, QR code factory, SVG converter, image upscaler, watermark add/remove, smart background remover.
• Fabric.js Graphic Canvas: Canva-like interactive graphic designer with layers, shapes, text styling, and export capabilities.
• Client-First & Edge Native: Heavy video/audio tasks execute client-side via WebAssembly (WASM) and Cloudflare Workers, keeping infrastructure costs at zero.

🤖 AI Agent CLI & MCP Suite (/cli)

• Codebase Intelligence (magetool inspect): Instant AST metrics, file distribution, and language analysis.
• Security & Secret Scanner (magetool audit): Detects exposed API keys (OpenAI, AWS), private keys, and code smells (eval(), prototype pollution, shell injections).
• Automated PR Reviewer (magetool review): Analyzes git diffs and generates GitHub-ready markdown code review comments with health scoring.
• Model Context Protocol Server (magetool serve): Native stdio MCP server for agentic IDEs.

🏗️ Architecture

graph TD
    User([User / AI Agent]) --> Gateway{MageTool Ecosystem}
    
    subgraph Web ["Web Media Suite (frontend/)"]
        Gateway -->|Browser UI| NextApp[Next.js App Router]
        NextApp --> WASM[Client-Side FFmpeg WASM]
        NextApp --> Canvas[Fabric.js Graphic Editor]
        NextApp --> EdgeWorker[Cloudflare Edge Services]
    end
    
    subgraph AI ["Developer CLI & MCP (cli/)"]
        Gateway -->|Terminal / RPC| CLI[MageTool CLI Router]
        CLI --> Scanner[Codebase Scanner]
        CLI --> Auditor[Security & Secret Auditor]
        CLI --> Reviewer[Automated PR Reviewer]
        CLI --> MCPServer[Stdio JSON-RPC MCP Server]
    end

🚀 Quickstart

1. Running the Web Platform

cd frontend
npm install
npm run dev

Open http://localhost:3000 to interact with the web studio.

2. Running the AI Agent CLI

cd cli
npm install
npm run build
node bin/magetool.js --help

🔌 Model Context Protocol (MCP) Integration

MageTool can be plugged into Claude Desktop, Cursor, Windsurf, or OpenAI Codex as an autonomous tool provider:

{
  "mcpServers": {
    "magetool": {
      "command": "node",
      "args": ["<path-to-magetool>/cli/bin/magetool.js", "serve"]
    }
  }
}

Provided Tools:

• magetool_inspect_repo: Inspects repo structure, file counts, and LOC.
• magetool_audit_security: Audits code for leaked credentials and security smells.
• magetool_pr_review: Generates markdown review comments for code diffs.

🧪 Testing & CI/CD

cd cli
npm test

MageTool uses automated GitHub Actions workflows for continuous integration testing and automated PR code reviews on every pull request.

🤝 Contributing

Contributions are warmly welcome! Please see CONTRIBUTING.md and CODE_OF_CONDUCT.md.

1. Fork the repo (git checkout -b feature/cool-feature)
2. Commit your changes (git commit -m 'feat: add cool feature')
3. Push to branch (git push origin feature/cool-feature)
4. Open a Pull Request

📄 License

Licensed under the Apache License, Version 2.0. See the LICENSE file for details.