🌌 Quark

The AI-Native, Cross-Platform Clipboard Micro-Daemon.

Quark is a headless, zero-configuration background service that orbits your operating system's clipboard. It silently fixes your copy-paste nightmares, syncs your clipboard across devices seamlessly, and exposes your clipboard directly to local AI models.

🛑 The Problem

The operating system clipboard is fundamentally broken and stuck in the past:

• Dumb Text: You copy text from a PDF, and the line breaks shatter.
• Privacy Leaks: You copy a link, and tracking tags (?utm_source=...) follow you.
• Lost Formatting: You copy a table or Markdown, and the formatting vanishes when pasted into an email.
• Isolation: You switch to your laptop, and your clipboard is empty.
• Blind AI: Your local AI agents (Claude, Cursor) have no native access to what you just copied.

💡 The Solution

Enter Quark. A silent, zero-configuration daemon that lives in your background. It requires no interface. It simply fixes what is broken.

• Heals broken PDF text automatically.
• Strips tracking parameters from URLs.
• Transforms raw Excel data, Markdown, and LaTeX into clean, rich HTML.
• Omni-Context Engine: Detects the active target app (Notion, Obsidian, VS Code, Pages) and dynamically shifts clipboard flavors (Markdown, HTML, RTF) to ensure flawless structure on every paste.
• Syncs across your Mac, Windows, and Linux machines instantly via P2P.
• Exposes your clipboard to local AI models via MCP.

🚀 Quick Start

No .exe files. No .dmg files. Just pure, cross-platform JavaScript.

# Run instantly without installing (Temporary Session)
npx -y @quark.clip/quark start

# OR install globally as a background OS service
npm install -g @quark.clip/quark
quark install

🏗️ Architecture

Quark is built entirely in Node.js to ensure 100% cross-platform compatibility without the need for native C++ compilation.

graph TD
    subgraph OS [Operating System]
        Clip[OS Clipboard]
    end

    subgraph Quark [Quark Daemon]
        Poll[Polling Engine]
        Trans[Smart Transformers]
        P2P[P2P Mesh Network]
        API[Local HTTP API]
    end

    subgraph Network [Local Wi-Fi]
        mDNS[mDNS Discovery]
        WS[WebSocket Sync]
    end

    subgraph AI [AI Agents]
        MCP[MCP Stdio Server]
        Claude[Claude Desktop / Cursor]
    end

    Clip <-->|Read/Write| Poll
    Poll -->|Raw Text/HTML| Trans
    Trans -->|Cleaned/Formatted| Poll
    Poll <-->|Broadcast/Receive| P2P
    P2P <--> mDNS
    P2P <--> WS
    API <--> Clip
    MCP <-->|HTTP GET/POST| API
    Claude <-->|Stdio| MCP

Technical Deep Dive

• Clipboard Engine: Uses native OS commands (pbcopy/pbpaste for macOS, xclip for Linux, powershell and System.Windows.Forms for Windows) to read and write both text/plain and text/html streams simultaneously.
• Zero-Config Mesh Network: Uses mDNS (Bonjour/ZeroConf) to broadcast its presence (quark-clip). When peers discover each other, they establish direct WebSocket connections. No central server is used.
• Transformation Pipeline: A synchronous/asynchronous pipeline that detects content types (Markdown, LaTeX, URLs, TSV) and applies regex heuristics, marked parsing, and smart typographic quoting based on language character frequency.

🤖 Model Context Protocol (MCP)

Quark acts as a bridge between your OS clipboard and local AI models using the Model Context Protocol.

When you run quark mcp, it starts a standard MCP server over stdio. This server communicates with the background Quark daemon via a local HTTP API (port 14314).

Exposed Tools

1. `get_clipboard`: Reads the user's current operating system clipboard.
2. `set_clipboard`: Writes text directly to the user's operating system clipboard.

Claude Desktop Integration

To give Claude Desktop access to your clipboard, add Quark to your claude_desktop_config.json:

{
  "mcpServers": {
    "quark": {
      "command": "npx",
      "args": ["-y", "@quark.clip/quark", "mcp"]
    }
  }
}

Note: Ensure the Quark daemon is running in the background (quark start or quark install) for the MCP server to function.

🤝 Contributing

Please read CONTRIBUTING.md for details on our code of conduct, and the process for submitting pull requests to us.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

© 2026 Adarsh Agrahari. All Rights Reserved.