桌面窗口捕获与视觉分析MCP服务器

desktop-mcp 是一个将 X11 窗口捕获与 OpenAI 视觉分析集成于一体的 MCP 服务器,可对桌面窗口截图进行智能识别与分析,适用于在开发工作流或自动化流程中快速获取窗口视觉信息的场景。

  • by AlanBlanchet
  • Python
  • 本地部署
  • 更新于
在 GitHub 查看 AlanBlanchet/desktop-mcp

使用说明

desktop-mcp

X11 window capture + OpenAI vision analysis as a single MCP server.

Tools

Tool Description
analyze_window Capture a window and analyze it with OpenAI vision

Installation

uv sync

Usage

// .vscode/mcp.json or ~/.config/Code/User/mcp.json
{
  "servers": {
    "desktop": {
      "type": "stdio",
      "command": "uv",
      "args": ["run", "desktop-mcp"],
      "env": {
        "OPENAI_API_KEY": "sk-...",
      },
    },
  },
}

Requirements

  • X11 (xwininfo)
  • maim
  • uv

Tests

uv run pytest tests/