kwin-mcp

概述

kwin-mcp 是一个全面的模型上下文协议（MCP）服务器，使 AI 代理能够与运行在 KDE Plasma 6 Wayland 上的 Linux 桌面应用程序交互。它既支持不影响用户桌面的隔离虚拟会话，也支持实时会话连接以实现协作工作流。该服务器提供 30 个 MCP 工具，涵盖鼠标、键盘、触摸、剪贴板、无障碍树检查、截图捕获和窗口管理，非常适合进行端到端 GUI 测试和桌面自动化。

试试问 AI

装完之后，这里有 5 个你可以让 AI 做的事：

• session_startapp_command? str, screen_width? int (1920), screen_height? int (1080), enable_clipboard? bool (false), keep_screenshots? bool (false), isolate_home? bool (false), keep_home? bool (false), env? dict

  Start an isolated KWin Wayland session, optionally launching an app

  试试:Start a KWin session with Firefox试试:Launch kcalc in an isolated session with screen size 1280x720
• session_connectdbus_address? str, wayland_display? str, keep_screenshots? bool (false)

  Connect to an existing KWin session (real desktop or container)

  试试:Connect to my existing KDE Plasma desktop试试:Connect to the KWin session in systemd-nspawn container
• screenshotinclude_cursor? bool (false)

  Capture a screenshot of the virtual display (saved as PNG)

  试试:Take a screenshot of the current desktop试试:Capture screenshot with cursor included
• accessibility_treeapp_name? str, max_depth? int (15), role? str

  Get the AT-SPI2 widget tree with roles, names, states, and coordinates

  试试:Get the accessibility tree for Firefox试试:Show all buttons in the current application
• find_ui_elementsquery str, app_name? str, states? list[str]

  Search for UI elements by name, role, or description with optional state filtering

  试试:Find the calculate button in kcalc试试:Search for all visible text input fields
• mouse_clickx int, y int, button? str ('left'), double? bool, triple? bool, modifiers? list[str], hold_ms? int (0), screenshot_after_ms? list[int]

  Click at coordinates with various options for button, modifiers, and click behavior

  试试:Click at position 100,200试试:Double-click the open button with Ctrl key held
• keyboard_typetext str, screenshot_after_ms? list[int]

  Type a string of text character by character (US QWERTY layout)

  试试:Type 'Hello World'试试:Enter the username 'admin' in the login field
• launch_appcommand str, env? dict

  Launch an application inside the running session

  试试:Launch Firefox browser试试:Start the calculator application
• list_windows

  List all accessible application windows with titles and focused state

  试试:Show all open windows试试:List applications with their window titles
• wait_for_elementquery str, app_name? str, timeout_ms? int (5000), poll_interval_ms? int (200), expected_states? list[str]

  Poll the accessibility tree until an element matching query and/or states appears

  试试:Wait for the login button to appear试试:Wait for the progress bar to become active
• mouse_dragfrom_x int, from_y int, to_x int, to_y int, button? str ('left'), modifiers? list[str], waypoints? list[[x,y,dwell_ms]], screenshot_after_ms? list[int]

  Drag from one point to another with smooth interpolation and optional waypoints

  试试:Drag file from position 100,200 to 500,600试试:Drag with waypoints to draw a custom shape
• focus_windowapp_name str

  Focus a window by application name

  试试:Focus the Firefox window试试:Bring the terminal application to front

可对比工具

安装

uv tool install kwin-mcp

pip install kwin-mcp

{
  "mcpServers": {
    "kwin-mcp": {
      "command": "uvx",
      "args": ["kwin-mcp"]
    }
  }
}

FAQ

kwin-mcp 支持哪些桌面环境？

kwin-mcp 专为 KDE Plasma 6 上的 Wayland 设计。它可以与任何在 KDE Plasma 上运行的 Wayland 应用程序协同工作，包括 Qt、GTK 和 Electron 应用。

kwin-mcp 如何处理输入而不触发系统授权提示？

kwin-mcp 直接使用 KWin 的私有 EIS（模拟输入服务器）D-Bus 接口，绕过了 XDG RemoteDesktop 门户，这消除了用户确认对话框的需要。

Hacker News 讨论

开发者社区最近的相关讨论。

• Show HN: Kwin-MCP – MCP server for AI-driven Linux GUI automation via KWin▲ 1
  帖子 by isac322 · 2026-02-23

kwin-mcp 对比