QuickDesk

QuickDesk is a remote desktop software positioned as "AI-native," featuring a deeply integrated MCP (Model Context Protocol) Server. This design allows AI clients like Cursor and Claude Desktop to directly capture remote screens, perform OCR, and execute over 40 desktop-level operations such as mouse clicks, keyboard input, and clipboard read/write.

AI Integration & MCP Protocol#

• Dual Transport Modes: Supports stdio mode (AI client spawns process) and HTTP/SSE mode (QuickDesk hosts server, allowing multi-client connections), with persistent selection across restarts
• 40+ MCP Tools: Screenshots (adjustable resolution for faster processing), mouse operations (click/drag/scroll), keyboard input (including shortcuts), clipboard read/write, OCR text recognition, UI element detection, screen verification, and more
• Host Skill Host: Built-in structured tools for system info, file operations, and shell execution; supports auto-loading and syncing of custom skills via directory addition
• Event-Driven Mechanism: Tools like wait_for_event and wait_for_screen_change for monitoring connection state, clipboard, screen changes, and real-time performance statistics
• MCP Resources & Prompts: Real-time device status resources, 9 built-in Prompt templates (remote operation guides, health checks, batch automation, system diagnostics, etc.)
• Background Automation: Supports show_window=false for headless batch operations

Remote Control & Transport#

• Chromium-Level Foundation: Based on Chromium Remoting protocol core, supporting H.264 / VP8 / VP9 / AV1 codecs
• WebRTC P2P Transport: ICE/STUN/TURN traversal, adaptive frame rate and bitrate, Office / Gaming frame rate boost modes
• Privacy Screen Mode: Physical display blackout and local input blocking during remote control (Windows 10 2004+)
• Interaction Sync: Full keyboard/mouse mapping, real-time remote cursor sync, bidirectional clipboard sync

Connection & Operations Management#

• Secure Connection: 9-digit Device ID + temporary access code (auto-refresh configurable: 30min to 24hr or never expire)
• Multi-Session Management: Multi-tab concurrent connections to multiple devices, connection history and quick reconnect
• Full-Link Performance Monitoring: Detailed latency breakdown panel (Capture → Encode → Network → Decode → Render), real-time FPS, bitrate, bandwidth, and RTT statistics

Architecture & Implementation#

The project utilizes a multi-language architecture: C++17 for the GUI and Chromium Remoting protocol core, Rust for the MCP Bridge and Skill Host, and Go for the signaling server. For performance optimization, it reuses highly optimized C++ code from Chromium and combines shared memory with Qt 6 QVideoSink to achieve a zero-copy pipeline from video capture to GPU rendering.

Core modules:

• QuickDesk/: Qt 6 GUI client with MVVM architecture and custom Fluent Design component library
• quickdesk-mcp/: Rust MCP Bridge handling stdio ↔ WebSocket protocol conversion
• quickdesk-skill-host/: Rust host skill host process (Cargo workspace)
• SignalingServer/: Go signaling server (based on Gin + GORM)
• WebClient/: Browser-based web client

For enterprise requirements, a complete self-hosted solution (Go signaling + PostgreSQL + Redis + coturn) is offered, with Docker-based quick deployment. Pre-built images published to ghcr.io/barry-ran/quickdesk-signaling.

Supported Platforms#

• Windows x64, macOS ARM64
• Linux and mobile platforms planned with no specific timeline

Unconfirmed Information#

• Author identity limited to GitHub user barry-ran, LICENSE credited to "Barry", no further public background
• Independent website and Telegram community links not provided with specific URLs in README
• Chromium Remoting code source unclear — whether directly referencing Chromium source or independently implementing a compatible protocol
• "The First AI-Native Remote Desktop" is a self-proclaimed marketing position without third-party validation