MCP ScreenCatch
Enables Claude Desktop to capture screenshots through an interactive region selection interface, automatically saving timestamped images to a configurable directory for analysis and documentation.
README
MCP ScreenCatch
An MCP (Model Context Protocol) server that provides screen capture functionality for Claude Desktop, emulating Windows Snipping Tool capabilities.
Overview
MCP ScreenCatch enables Claude Desktop to capture screenshots of user-selected screen regions. The tool allows users to highlight areas of their monitor(s), capture them, and save the images with timestamps for easy retrieval and processing.
Objectives
- Seamless Integration: Provide an MCP command that Claude Desktop can invoke to initiate screen captures
- Interactive Selection: Allow users to select specific regions of their screen(s) using a familiar snipping interface
- Multi-Monitor Support: Work across multiple monitors for flexible screen capture
- Sequential Captures: Enable users to capture multiple screenshots in succession without restarting the tool
- Timestamped Files: Automatically timestamp each capture for chronological ordering
- Configurable Output: Let users specify the output directory for captured images
- Pipeline Ready: Save files in a location accessible to other MCP commands or external applications
Features
Core Functionality
- Region Selection: Click and drag interface to select screen regions
- Capture & Save: Button/icon to finalize and save the selected region
- Continuous Mode: After each capture, prompt user to:
- Capture another screenshot
- Quit the tool
- Timestamp Naming: Files named with ISO 8601 timestamps (e.g.,
capture_2025-11-27_064730.png) - Directory Selection: User-configurable output directory
Use Cases
- Capture UI elements for debugging
- Document visual workflows
- Create sequential tutorials or guides
- Feed captured images to Claude for analysis
- Build visual documentation pipelines
Technical Implementation
MCP Protocol
The tool will be exposed as an MCP server that Claude Desktop can connect to, providing:
- Tool:
capture_screen- Initiates the screen capture interface - Tool:
set_output_directory- Configure where captures are saved - Tool:
list_captures- Retrieve timestamped capture files
Technology Stack
- Language: TypeScript/Node.js
- MCP SDK: @modelcontextprotocol/sdk
- Screen Capture: screenshot-desktop for cross-platform screen capture
- Image Processing: Sharp for image cropping and manipulation
- UI Framework: Electron for transparent overlay and region selection
File Format
- Default format: PNG (lossless, good for UI captures)
- Naming convention:
capture_YYYY-MM-DD_HHMMSS.png - Metadata: Optional JSON sidecar with capture timestamp and dimensions
Installation
npm install
npm run build
Configuration
Add to Claude Desktop config (claude_desktop_config.json):
{
"mcpServers": {
"screencatch": {
"command": "node",
"args": ["/path/to/mcp-screencatch/build/index.js"]
}
}
}
Usage
From Claude Desktop
Once configured, you can ask Claude to capture screenshots:
Claude, can you capture a screenshot of my screen?
Interactive Overlay
- An Electron overlay window will appear covering your screen(s)
- Click and drag to select the region you want to capture
- Click the Capture button or press Enter to save
- Press ESC or click Cancel to abort
- The captured image is saved with a timestamp
Available Commands
Capture a screenshot:
Claude, capture a screenshot
Set output directory:
Claude, set the screenshot output directory to C:\Users\johne\screenshots
List recent captures:
Claude, list my recent screenshots
File Naming
Screenshots are automatically named with ISO 8601 timestamps:
capture_2025-11-27_143052.pngcapture_2025-11-27_143105.png
This ensures chronological ordering and prevents filename conflicts.
Development Status
✅ Implemented - Core functionality complete
- ✅ MCP server with stdio transport
- ✅ Interactive region selection with Electron overlay
- ✅ Timestamped file naming
- ✅ Configurable output directory
- ✅ List and manage captures
- ⚠️ Note: Requires Electron, which adds ~150MB to node_modules. For production use, consider electron-builder for packaging.
Future Enhancements
- Support for full screen and window captures
- Multiple image format options (PNG, JPG, WebP)
- Annotation tools (arrows, text, highlights)
- Clipboard integration
- Automatic cleanup of old captures
- Screenshot comparison tools
License
MIT
Contributing
Contributions welcome! Please open an issue or submit a pull request.
推荐服务器
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。
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。
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。
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。
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。
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。
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。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
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。