By lostsunset
Give Claude Code eyes beyond the browser — real-time desktop screenshot capture, streaming, and UI analysis via MCP
Give Claude Code eyes beyond the browser.
DesktopLens MCP 是一個 Model Context Protocol (MCP) Server,讓 Claude Code 能即時捕獲桌面應用程式的 UI 畫面、進行即時串流、UI 比較分析,並支援 Plugin 擴充生態系統。
| 工具 | 說明 |
|---|---|
desktoplens_list_windows | 列出所有可見桌面視窗,支援模糊搜尋 |
desktoplens_screenshot | 截取指定視窗截圖,自動儲存快照,支援品質調整與 grid overlay 標注 |
desktoplens_status | 回傳伺服器狀態、平台資訊、串流 session 資訊 |
desktoplens_watch | 開始即時串流指定視窗,透過 WebSocket 傳輸至 Chrome Viewer |
desktoplens_stop | 停止串流 session (指定或全部) |
desktoplens_compare | 比較兩張截圖差異 — 像素比對、變化區域偵測、diff image 生成 |
| 工具 | 說明 |
|---|---|
desktoplens_plugin_search | 在 GitHub marketplace 搜尋 DesktopLens Plugin |
desktoplens_plugin_install | 從本地路徑安裝 Plugin |
desktoplens_plugin_list | 列出所有已安裝的 Plugin |
desktoplens_plugin_remove | 移除已安裝的 Plugin |
透過 Claude Code Plugin Marketplace 一鍵安裝,自動啟動 MCP Server 並載入所有 skills、commands、agents:
/plugin marketplace add LostSunset/DesktopLens_MCP
/plugin install desktoplens-mcp@desktoplens-mcp
安裝後可直接使用:
/screenshot [window] — 快速截取桌面視窗截圖/windows [filter] — 列出所有可見視窗/watch [window] — 開始即時串流或者,將以下設定手動加入你的 MCP 設定:
{
"mcpServers": {
"desktoplens": {
"command": "npx",
"args": ["-y", "desktoplens-mcp"]
}
}
}
> 列出我桌面上所有視窗
> 截取 Notepad 的畫面並分析 UI 佈局
> 開始即時串流 VS Code 的畫面
> 比較截圖前後的 UI 差異
> 搜尋可用的 DesktopLens Plugin
> 停止所有串流 session
> 使用 desktoplens_watch 工具串流 "Visual Studio Code" 視窗
回傳:
{
"session_id": "abc-123",
"stream_url": "ws://localhost:9876/stream/abc-123",
"viewer_url": "http://localhost:9876/?session=abc-123",
"status": "streaming"
}
串流支援自適應 FPS (0.5-5fps)、三種品質級別 (low/medium/high)、dirty block 差異壓縮。
> 搜尋 UI 相關的 Plugin
> 安裝 Plugin: /path/to/my-plugin
> 列出已安裝的 Plugin
Plugin 使用命名空間隔離:plugin_{pluginName}_{toolName}。詳見 Plugin 開發指南。
Claude Code ◄─ stdio MCP ─► MCP Server (10 tools)
│
┌────────────────┼────────────────┐
│ │ │
Capture WebSocket Plugin
Engine Streaming System
│ │ │
Screenshots Chrome Viewer Marketplace
詳見 系統架構文件。
# 安裝依賴
npm install
# TypeScript 型別檢查
npx tsc --noEmit
# 開發模式
npm run dev
# 執行測試 (329 tests)
npm test
# 執行測試 + 覆蓋率 (100%)
npm run test:coverage
# 編譯
npm run build
| 平台 | 截圖方式 | 狀態 |
|---|---|---|
| Windows | node-screenshots (Rust native) | ✅ 完整支援 |
| macOS | node-screenshots (Rust native) | ✅ 完整支援 |
| Linux | node-screenshots (Rust native) | ✅ 完整支援 |
| 變數 | 預設值 | 說明 |
|---|---|---|
DESKTOPLENS_PORT | 9876 | WebSocket 埠號 |
DESKTOPLENS_DEFAULT_FPS | 2 | 預設串流幀率 |
DESKTOPLENS_DEFAULT_QUALITY | medium | 預設畫質 (low/medium/high) |
DESKTOPLENS_PLUGIN_DIR | ~/.desktoplens/plugins | Plugin 目錄 |
DESKTOPLENS_LOG_LEVEL | info | 日誌等級 (debug/info/warn/error) |
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
npx claudepluginhub lostsunset/desktoplens_mcp --plugin desktoplens-mcpPDF to PNG converter and academic paper tools for Claude Code
Use when the user explicitly asks for a desktop or system screenshot (full screen, specific app or window, or a pixel region), or when tool-specific capture capabilities are unavailable and an OS-level capture is needed. Originally from OpenAI's curated skills catalog.
Give Claude eyes and hands — screen capture and interaction for full-auto workflows
Visual feedback capture for macOS app windows.
Computer use toolkit for driving desktop environments through Claude's vision and action API with screenshot capture, mouse/keyboard control, and an autonomous agent loop
Safe cross-OS desktop & browser control for Claude Code — the fallback execution layer for when APIs, CLIs, and direct integrations aren't available. Registers the clawdcursor MCP server in compact mode and bundles its usage skill (the SKILL.md at the repo root). Requires Node.js 20+; the MCP server launches via npx, which fetches clawdcursor on demand (or uses your global install if present) — no separate install step.
MCP server + skill for Linux KDE Plasma 6 Wayland GUI automation. Provides 30 MCP tools (mouse, keyboard, touch, clipboard, screenshot, AT-SPI2 accessibility tree, window mgmt, D-Bus passthrough) plus the kwin-desktop-automation skill that guides session-mode selection, observation/action sequencing, and platform pitfalls.