mcp-youtube-transcript

This MCP server provides tools to fetch YouTube video transcripts and metadata. The core tools are: get_transcript, which returns the full transcript for a given YouTube URL; get_timed_transcript, which returns the transcript with timestamps; and get_video_info, which fetches metadata about the specified video. Transcripts can be paginated when they exceed token limits; you may receive a next_cursor to retrieve the next chunk. To use the server, deploy it via uvx and point requests to the youtube-transcript MCP server name. When querying, pass the required URL parameter and optional language (default en) and cursor to navigate large transcripts. If you need to adjust how the response is chunked, you can use the --response-limit option to control the maximum number of characters per response. This server also supports proxy configuration for restricted networks via standard environment variables or command-line proxy options.

Prerequisites:

• Python and uv (Astral UV) installed on your system. See the uv documentation for installation details.
• Access to Git (for the from git installation method).
• Optional: Docker if you prefer running via a container.

Installation steps (via uvx from Git):

1. Install uv as per the project’s guidance (this may involve following Astral UV’s install instructions).
2. Install Git if you don’t have it already.
3. Run the MCP server install using uvx from the repository:

# Install from the Git source (the recommended approach in the README)
uvx --from git+https://github.com/jkawamoto/mcp-youtube-transcript mcp-youtube-transcript

4. Start the MCP server (the exact command may vary depending on your environment; typically you would launch the UV-based runtime with the installed MCP bundle).
5. Verify the server is reachable and the tools (get_transcript, get_timed_transcript, get_video_info) respond to requests.

Alternative: Docker

• Open the Docker Hub page for the youtube_transcript MCP and follow the usage instructions there to run the image directly. This typically involves pulling the image and running it with appropriate port mappings and environment variables.

Tips and considerations:

• If transcripts exceed typical token limits, use the --response-limit option to split responses into chunks (e.g., 15000 or 50000 characters) and rely on next_cursor for continuation.
• If you operate in a network that restricts YouTube access, configure proxies via HTTP_PROXY/HTTPS_PROXY or the Webshare proxy options described in the README.
• For local Claude Desktop or LM Studio integrations, you can configure the MCP server entry using the provided json snippets in the README to connect to this youTube transcript server.
• The server supports language selection through the lang parameter (default en). Ensure the language code you pass is supported by the transcripts API you’re using.
• In Kubernetes or other orchestrators, you can expose the uvx-based service and route requests to the youtube-transcript MCP server name.