Elgato MCP Server
Bridges AI assistants with Elgato apps via IPC, enabling dynamic tool discovery and notification forwarding.
README
Elgato MCP Server
A Model Context Protocol (MCP) server that bridges AI assistants (like Claude Desktop) with Elgato apps.
Overview
The Elgato MCP Server acts as a protocol bridge between MCP clients and Elgato apps via IPC:
MCP Client <--MCP Transport--> Bridge <--Unix Socket/Named Pipe--> Elgato App
Key Features:
- 🔌 Dynamic Tool Discovery — Automatically discovers and exposes tools from connected Elgato apps via MCP
- 🚀 Dual Transport Support — stdio (for Claude Desktop) and HTTP (for web clients)
- 🌐 ngrok Integration — Optional public tunnel for remote access
- 🔄 Hot Reconnection — Automatically reconnects when apps become available
- 💻 Cross-Platform — Supports Windows and macOS
- 📢 Notification Forwarding — Forwards app notifications to connected MCP clients
Installation
# Global installation (recommended)
npm install -g @elgato/mcp-server
# Or with pnpm
pnpm add -g @elgato/mcp-server
Running without Installation
You can run the Elgato MCP Server directly without installing it globally using npx:
# Run with stdio transport (default)
npx -y @elgato/mcp-server@latest
# Run with HTTP transport
npx -y @elgato/mcp-server@latest --http
# Run with HTTP transport on a custom port
npx -y @elgato/mcp-server@latest --http --port 3000
# Run with ngrok tunnel
NGROK_AUTHTOKEN=your_token npx -y @elgato/mcp-server@latest --http --ngrok
# Run with verbose logging
npx -y @elgato/mcp-server@latest --verbose
This is useful for trying out the server without a permanent installation, or for use in CI/CD environments.
Usage
stdio Transport (Default)
For integration with Claude Desktop or other MCP clients using standard I/O:
elgato-mcp-server
HTTP Transport
For web-based clients or remote access:
# Start HTTP server on default port (9090)
elgato-mcp-server --http
# Custom port
elgato-mcp-server --http --port 3000
# With ngrok tunnel (requires NGROK_AUTHTOKEN env var)
NGROK_AUTHTOKEN=your_token elgato-mcp-server --http --ngrok
CLI Options
Options:
--transport <mode> Transport mode: 'stdio' (default) or 'http'
--http Shorthand for --transport http
--port <number> HTTP server port (default: 9090)
--ngrok Enable ngrok tunnel (requires NGROK_AUTHTOKEN env var)
--help, -h Show help message
--verbose, -v Enable verbose logging (default: silent)
Logging
Logs are written to stderr with a severity label (ERROR, WARN, INFO, DEBUG). Errors and warnings always output; info/debug require --verbose.
Claude Desktop Configuration
Add the following to your Claude Desktop configuration file:
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
Using Global Installation
{
"mcpServers": {
"elgato": {
"command": "elgato-mcp-server"
}
}
}
Using npx (No Installation Required)
Alternatively, you can use npx to run the server without installing it globally:
{
"mcpServers": {
"elgato": {
"command": "npx",
"args": ["--yes", "@elgato/mcp-server@latest"]
}
}
}
Note: The npx-based configuration works with Claude Desktop and other MCP clients that support the stdio transport. The first run may take a moment while npx downloads the package.
HTTP Endpoints
When running in HTTP mode, the following endpoints are available:
| Endpoint | Method | Description |
|---|---|---|
/mcp |
POST | MCP request endpoint |
/mcp |
GET | Server-Sent Events (SSE) for streaming |
/mcp |
DELETE | Close session |
/health |
GET | Health check endpoint |
Requirements
- Elgato app with MCP plugin support (e.g. Stream Deck)
- Node.js 18 or later
- Supported platforms: Windows, macOS
Contributing
We welcome contributions! For development setup, coding guidelines, and the contribution process, see CONTRIBUTING.md.
For detailed technical documentation and architecture information, see TECHNICAL_SPECIFICATION.md.
License
MIT License - Copyright (c) Corsair Memory Inc.
See LICENSE for details.
Links
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。