Windows Context MCP 🪟⚡

Windows Context MCP is a high-performance, native Windows service, Model Context Protocol (MCP) server, and Unified Personal Context Hub connecting Windows desktop activity and Android mobile telemetry directly into conversational AI assistants like Google Gemini, OpenAI ChatGPT, and Claude.

🌟 Features & Architecture

• 🗄️ Zero-Dependency SQLite Time-Series (src/db.ts): Built on native node:sqlite for persistent indexed focus sessions, 24-hour hourly distributions, and multi-day range analytics.

• 📱 Unified Cross-Device Hub (src/aggregator.ts): Ingests Android screen time and telemetry via /api/sync, calculating combined multi-device screen time, cross-device top applications, and device fleet health.

• 🔄 Dual Protocol Transports:

  • Streamable HTTP / JSON-RPC 2.0 (src/server.ts): Remote MCP gateway on port 3001 for Google Gemini & ChatGPT Custom GPT Actions.

  • Native STDIO Transport (src/stdio.ts): Standard I/O stream for Claude Desktop, Cursor, and local CLI agents (node dist/index.js --stdio).

• 🪟 Native Win32 API Queries (src/collector.ts): Direct User32 / CIM queries for active window titles, process binaries, CPU delta utilization, RAM usage, user idle detection, and battery telemetry.

• 🧪 Comprehensive Test Suite: 23 automated unit and integration tests across SQLite database, cross-device aggregation, and tool executions (npm test).

Related MCP server: system-media-control-mcp

🛠️ MCP Tools Reference (15 Tools)

🚀 Quick Start

1. Run via NPX (1-Click Execution)

npx windows-context-mcp --stdio

2. Build & Test Locally

npm install
npm test
npm run build
npm start

HTTP server starts on port 3001 with endpoint http://localhost:3001/mcp and live UI at http://localhost:3001.

🔗 AI Assistant Setup

Claude Desktop (%APPDATA%\Claude\claude_desktop_config.json)

{
  "mcpServers": {
    "windowsContext": {
      "command": "node",
      "args": ["C:/Users/smnk2/.gemini/antigravity/scratch/windows-context-mcp/dist/index.js", "--stdio"]
    }
  }
}

Google Gemini / ChatGPT (HTTP Mode)

{
  "mcpServers": {
    "windowsContext": {
      "url": "http://localhost:3001/mcp",
      "transport": "http"
    }
  }
}

👨‍💻 Author & Maintainer

• Nandhakumar Murugan (@nandhakumar-murugan)

• Google Student Ambassador 🌐 | B.E. CS & Cyber Security @ KG KiTE 🛡️

• Founder @ Prema AI Labs 🚀

📄 License

This project is licensed under the MIT License.