pov

pov

Enables LLM agents to capture screenshots, control mouse/keyboard, and manage windows on desktop platforms, primarily Windows, via an MCP server.

Category
访问服务器

README

pov - Point of View

Screenshot capture, mouse/keyboard control, and window management — as a CLI and MCP server for LLM vision agents.

Lets LLMs (Claude Desktop, Cursor, OpenCode) see and interact with the Windows desktop. Works on Windows, macOS, Linux (X11/Wayland), and WSL.

Install

# With uv
uv pip install pov-desktop

# With pip
pip install pov-desktop

To install the pov CLI as a standalone tool:

uv tool install pov-desktop

Or from source:

git clone https://github.com/empezarcero/pov.git
cd pov
uv sync

CLI Usage

# Screenshots
pov capture                     # Capture screenshot, print base64
pov capture --output shot.png   # Save to file
pov capture --monitor 1         # Specific monitor
pov monitors                    # List available monitors

# Mouse
pov click                       # Left-click at current position
pov click 500 300               # Left-click at (500, 300)
pov click 500 300 --button right --clicks 2
pov move-mouse 800 600          # Move cursor to (800, 600)
pov scroll -3                   # Scroll down 3 clicks
pov scroll 5 --x 400 --y 200   # Scroll up at position
pov cursor                      # Print current cursor position

# Keyboard
pov type "Hello, world!"        # Type text literally
pov key enter                   # Press a key
pov key ctrl+c                  # Key combo
pov key alt+f4                  # Another combo

# Windows
pov windows                     # List visible windows
pov processes                   # List windowed processes
pov foreground                  # Get the focused window
pov focus <hwnd>                # Focus a window by handle
pov window-state <hwnd> minimize
pov move-window <hwnd> --x 100 --y 100
pov resize-window <hwnd> --width 1024 --height 768
pov close-window <hwnd>         # Gracefully close a window

# MCP server
pov serve                       # stdio transport (default)
pov serve --transport sse --port 8000

MCP Server

The MCP server exposes 16 tools:

Screenshots

Tool Description
screenshot Capture a screenshot and return it as an image
list_monitors List available monitors and their geometry

Mouse

Tool Description
mouse_move Move the cursor to screen coordinates
mouse_click Click at coordinates (or current position)
mouse_scroll Scroll the mouse wheel
get_cursor_position Get the current cursor position

Keyboard

Tool Description
keyboard_type Type a string of text literally
keyboard_key Press a key or combo (e.g. ctrl+c, alt+f4)

Window Management

Tool Description
list_windows List all visible windows with handles, titles, positions
focus_window Bring a window to the foreground
set_window_state Minimize, maximize, restore, hide, or show a window
move_window Move and/or resize a window
resize_window Resize a window without moving it
get_foreground_window Get info about the currently focused window
close_window Gracefully close a window (WM_CLOSE)
list_processes List running processes with visible windows

Claude Desktop / Cursor / OpenCode

Add to your MCP config:

{
  "mcpServers": {
    "pov": {
      "command": "uv",
      "args": ["run", "--directory", "/path/to/pov", "pov", "serve"]
    }
  }
}

SSE Transport

{
  "mcpServers": {
    "pov": {
      "url": "http://127.0.0.1:8000/sse"
    }
  }
}

Platform Support

Platform Backend Notes
Windows mss + ctypes Screenshots via mss, input/windows via Win32
macOS mss Screenshot capture only
Linux mss Screenshot capture only (X11 and Wayland)
WSL PowerShell + .NET/Win32 Full support — captures the Windows desktop, controls mouse/keyboard, manages windows

Development

# Install in development mode
uv sync

# Run the CLI
uv run pov --help

# Run the MCP server
uv run pov serve

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选