一个简单的MCP服务器，让大型语言模型轻松获取YouTube视频的文字稿。

安装

把这个添加到你的LLM应用的MCP设置里：

如果你更喜欢节点而不是bun，可以用npx代替：

用

只要问问你的LLM关于YouTube视频的建议：

总结本视频：https://www.youtube.com/watch?v=dQw4w9WgXcQ

这个视频的三个关键点是什么？https://www.youtube.com/watch?v=dQw4w9WgXcQ

这里对这个论点有哪些反驳？https://www.youtube.com/watch?v=dQw4w9WgXcQ

youtube-transcript-mcp

A simple MCP server that lets LLMs easily get the transcript of YouTube videos.

To install

Add this to your LLM app’s MCP settings:

{
  "mcpServers": {
    "youtube-transcript": {
      "command": "bunx",
      "args": ["--bun", "-y", "youtube-transcript-mcp"]
    }
  }
}

If you prefer node to bun, you can use npx instead:

{
  "mcpServers": {
    "youtube-transcript": {
      "command": "npx",
      "args": ["-y", "youtube-transcript-mcp"]
    }
  }
}

To use

Just ask your LLM about a YouTube video:

Summarize this video: https://www.youtube.com/watch?v=dQw4w9WgXcQ

What are the three key points of this video? https://www.youtube.com/watch?v=dQw4w9WgXcQ

What are some counterpoints to this argument here? https://www.youtube.com/watch?v=dQw4w9WgXcQ