System: Control your Mac from anywhere using natural language

Summary

SYSTEM remote mac automation Control your Mac from anywhere using natural language. Built with Cloudflare Agents SDK for intelligent scheduling, memory, and tool orchestration. agent url https://your-agent.workers.dev quick start 1. clone and install git clone https://github.com/ygwyg/system cd system && npm install 2. run setup wizard npm run setup Interactive setup: Anthropic API key, Raycast extensions, remote access. 3. start system npm start Starts bridge, tunnel, and opens the agent UI. architecture SYSTEM uses a split architecture for security: the Agent (brain) runs on Cloudflare Workers, while the Bridge (body) runs locally on your Mac. ┌───────────────────────────────────────────────────────┐ │ USER │ │ (phone/browser) │ └─────────────────────┬─────────────────────────────────┘ │ HTTPS ▼ ┌───────────────────────────────────────────────────────┐ │ AGENT (Brain) │ │ Cloudflare Workers │ │ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ │ │ │ Claude │ │ State │ │ Schedules │ │ │ │ AI │ │ (D.O.) │ │ (D.O.) │ │ │ └─────────────┘ └─────────────┘ └─────────────┘ │ └─────────────────────┬─────────────────────────────────┘ │ Tunnel ▼ ┌───────────────────────────────────────────────────────┐ │ BRIDGE (Body) │ │ Your Mac (local) │ │ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ │ │ │ AppleScript │ │ Shell │ │ Raycast │ │ │ └─────────────┘ └─────────────┘ └─────────────┘ │ └───────────────────────────────────────────────────────┘ agent (brain) • Cloudflare Workers + D.O. • Claude for NLP • State, memory, scheduling • WebSocket real-time bridge (body) • Local Express server • AppleScript, shell exec • Raycast extensions • Cloudflare Tunnel authentication All requests require an API secret via Bearer token or query parameter. Authorization: Bearer <api_secret> ?token=<api_secret> note The API secret is generated during npm run setup and stored in bridge.config.json. chat Send natural language commands to control your Mac. Send a message to the agent. request { "messag...