jkawamoto/mcp-youtube-transcript

What it solves

This project provides a way for AI agents and LLMs to programmatically retrieve transcripts and metadata from YouTube videos, allowing them to "read" the content of a video without needing to watch it.

How it works

It is implemented as a Model Context Protocol (MCP) server. This means it acts as a bridge between an LLM (like Claude or LM Studio) and the YouTube Transcript API. It provides a set of tools that the LLM can call to fetch transcripts (with or without timestamps), check available languages, and retrieve video metadata.

Who it’s for

Developers and users of AI agents (such as Goose, Claude Desktop, or LM Studio) who want their AI to be able to process and analyze YouTube video content.

Highlights

• Pagination Support: Automatically splits transcripts longer than 50,000 characters (adjustable via --response-limit) to avoid exceeding LLM token limits.
• Timed Transcripts: Ability to fetch transcripts with precise timestamps for better referencing.
• Language Selection: Support for specifying the desired transcript language.
• Proxy Support: Includes built-in support for residential proxies (like Webshare) and standard HTTP/HTTPS proxies to bypass IP bans.