
Claude Desktop
桌面端 · 免费增值 · Proprietary
Anthropic 官方推出的 Claude AI 桌面应用程序。支持 MCP 服务器以扩展功能。

jasondsmith72/tpc-mcpv2
一个基于nut.js的MCP服务器,提供对电脑屏幕、鼠标和键盘的全面控制功能,包括截图、鼠标操作、键盘输入、窗口管理和剪贴板访问等。
npm install -g cmake-js{
"mcpServers": {
"total-pc-control": {
"command": "node",
"args": [
"/ABSOLUTE/PATH/TO/total-pc-control/build/index.js"
]
}
}
}capture_screen
Capture the entire screen as an image.
capture_region
Capture a specific region of the screen.
get_screen_size
Get the dimensions (width and height) of the screen.
move_mouse
Move the mouse cursor to a specific x, y position.
get_mouse_position
Get the current x, y position of the mouse cursor.
click_mouse
Click the mouse at the current position.
click_at
Click the mouse at a specific x, y position.
double_click
Double-click the mouse at the current position.
double_click_at
Double-click the mouse at a specific x, y position.
scroll_mouse
Scroll the mouse wheel.
drag_mouse
Drag the mouse from the current position to a target x, y position.
drag_mouse_from_to
Drag the mouse from a startX, startY position to an endX, endY position.
type_text
Type text at the current cursor position.
type_text_with_delay
Type text with a delay between keystrokes.
press_key
Press a specific keyboard key.
press_key_shortcut
Press a keyboard shortcut (combination of keys).
hold_key
Hold down a keyboard key.
release_key
Release a held keyboard key.
get_clipboard_text
Get text from the clipboard.
set_clipboard_text
Set text to the clipboard.
copy_selected_text
Copy selected text to clipboard and return it (uses Ctrl+C/Cmd+C).
paste_text
Paste text at current cursor position (uses Ctrl+V/Cmd+V).
get_clipboard_image
Get image from the clipboard (if available) as base64 data.
get_ui_element_info
Finds a UI element within a specified window and returns its properties (Name, AutomationId, ClassName, ControlType, BoundingRectangle, IsEnabled, IsOffscreen,…
invoke_ui_element_action
Performs an action on a specified UI element.
本页信息摘自该项目的 README。 查看 README
该服务器 README 中提到的客户端:

microsoft/playwright
Playwright是一个跨浏览器的Web测试和自动化框架,支持Chromium、Firefox和WebKit,提供无头执行、多浏览器兼容性测试、强大的断言和工具支持,适用于多种编程语言。

yamadashy/repomix
Repomix是一个将代码库打包为AI友好格式的工具,支持本地和远程仓库处理,提供代码压缩、安全检查和多种输出格式。

bytedance/UI-TARS-desktop
TARS 是字节跳动的多模态 AI 智能体技术栈,包含两个项目:Agent TARS(基于 MCP 的命令行与 Web UI 智能体)和 UI-TARS-desktop(桌面 GUI 智能体)。

ahujasid/blender-mcp
BlenderMCP通过MCP协议将Blender与Claude AI连接,实现AI辅助3D建模与场景控制

microsoft/playwright-mcp
Playwright MCP是一个基于Playwright的浏览器自动化服务器,通过结构化数据而非像素输入实现LLM与网页的交互。

comet-ml/opik
Opik是一个开源的LLM评估框架,支持追踪、评估和监控LLM应用,帮助开发者构建更高效、更经济的LLM系统。