The YouTube Transcript MCP is a practical tool designed to pull written text directly from YouTube videos. Instead of having to watch a long video or manually transcribe its content, this server allows an AI to "read" exactly what was said. By simply providing a video link or ID, the tool retrieves the captions or subtitles, making it easy to summarize long presentations, find specific quotes, or translate video dialogue into different languages without leaving the chat interface. Beneath its simple interface, the server provides a robust tool called `get_transcript` that handles various URL formats and video identifiers. It offers significant flexibility by allowing users to specify language codes, such as English or Korean, ensuring the AI captures the correct subtitle track. The system is built with resilience in mind, featuring sophisticated error handling for scenarios like private videos, missing transcripts, or network interruptions, which ensures a reliable connection between the video platform and the AI. For developers building AI-powered workflows, this MCP server serves as a critical bridge that converts unstructured video data into clean, structured text that large language models (LLMs) can process efficiently. It can be easily integrated into Claude Desktop or other MCP-compatible clients via Smithery or a simple JSON configuration. This enables complex automated tasks, such as generating markdown documentation from video tutorials, performing sentiment analysis across video series, or creating searchable databases from video-based educational content.
Category: Design, Media & Creative
Tags: captions, subtitles, transcription, video, youtube
bash npx -y @smithery/cli install @kimtaeyoon83/mcp-server-youtube-transcript --client claude Via mcp-get: shell npx @michaellatman/mcp-get@latest install @kimtaeyoon83/mcp-server-youtube-transcript Manual Development Setup: 1. Install dependencies: npm install 2. Build the server: npm run build ---json { "mcpServers": { "youtube-transcript": { "command": "npx", "args": ["-y", "@kimtaeyoon83/mcp-server-youtube-transcript"] } } } ---url (string, required): YouTube video URL or video ID. * lang (string, optional, default: "en"): Language code for transcript (e.g., 'en', 'ko'). ---get_transcript tool, the AI can pull the instructional dialogue and technical steps. It can then format this raw text into structured Markdown documentation, complete with code blocks and headings. Example: "Fetch the transcript for this React tutorial [URL] and turn it into a step-by-step technical blog post with clear instructions."X, LinkedIn, or Newsletters), but rewriting video scripts into short-form text takes significant manual effort. Solution: The MCP retrieves the video's spoken content, which the LLM can then rewrite into different formats—such as a viral Twitter thread, a professional LinkedIn post, or a newsletter blurb—while maintaining the original voice of the creator. Example: "Extract the transcript from my latest YouTube upload [URL] and draft 5 engaging tweets highlighting the most controversial points discussed."
lang parameter to retrieve transcripts in the original language. The LLM can then perform a high-quality translation or provide a cross-lingual analysis of the content. Example: "Get the Korean transcript for this tech review [URL] and translate the specific criticisms of the hardware into English for my market research report."Part of MCP Servers
You can install the server for Claude Desktop automatically by running npx -y @smithery/cli install @kimtaeyoon83/mcp-server-youtube-transcript --client claude in your terminal. Alternatively, you can install it using mcp-get or add the npx command configuration directly to your client settings file.
The server retrieves complete captions and subtitles from YouTube video URLs, Shorts, and video IDs. It supports language selection with automatic fallbacks, optional timestamp inclusion for exact timing, and built-in filtering of advertisements or sponsored segments. It also provides an optional video analysis tool using TwelveLabs Pegasus.
No, the transcript retrieval tool operates without any YouTube API keys or external authentication credentials. You can fetch public captions immediately after setup. An API key is only required if you choose to enable the optional TwelveLabs Pegasus integration for direct video visual analysis.
Any Model Context Protocol client supporting stdio communication can run YouTube Transcript MCP. It is pre-configured and tested with Claude Desktop, and can be integrated into custom agent workflows, terminal MCP runners, and other compatible developer environments supporting standard MCP server definitions.
Yes, YouTube Transcript MCP is open source software distributed under the permissive MIT License. Users and developers can review the TypeScript codebase, report issues, submit pull requests, or adapt the server configuration for custom automated workflows directly on the project's official GitHub repository.