Skip to main content
Glama
YouTube Transcript MCP Server

YouTube Transcript MCP Server

by jkawamoto

English | 日本語

YouTube Transcript MCP Server

uv Python Application pre-commit GitHub License Dockerhub

A Model Context Protocol (MCP) server that fetches transcripts and metadata from YouTube videos directly into your LLM workflows.

Designed specifically for AI agents, it seamlessly handles long-form videos through automatic pagination and provides robust proxy support to circumvent YouTube rate limits and IP bans.

mcp-youtube-transcript

The header image was generated with Muse (Muse Image).

Key Features

  • Rich Transcript Retrieval: Fetch raw text, timestamped segments, and video metadata in multiple languages.

  • Smart Chunking & Pagination: Automatically chunks long transcripts (default: 50,000 characters) to prevent context window overflow, letting models page through hours of video effortlessly.

  • Resilient Proxy Support: Built-in support for residential proxies (Webshare, ScrapingAnt) and standard HTTP/HTTPS proxies to prevent IP blocks.

  • Universal Compatibility: Works with Claude Desktop, Cursor, LM Studio, Goose, and any standard MCP client.

Related MCP server: YouTube Transcript Server

Quick Start & Installation

NOTE

You'll needuv installed on your system to use uvx command.

This server communicates via standard stdio. Most MCP clients can run it directly using uvx (part of Astral uv).

General Configuration (Standard MCP JSON)

Add this entry to your client's MCP configuration file (typically under mcpServers):

{
  "mcpServers": {
    "youtube-transcript": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/jkawamoto/mcp-youtube-transcript",
        "mcp-youtube-transcript"
      ]
    }
  }
}

Claude Desktop

  • GUI (Drag & Drop): Download the .mcpb bundle from the Releases page and drop it into your Claude Desktop Settings.

  • Manual Config: Add the JSON above to claude_desktop_config.json (restart Claude Desktop after saving):

    • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

    • Windows: %APPDATA%\Claude\claude_desktop_config.json

Cursor

  1. Go to Cursor Settings > Features > MCP Servers.

  2. Click + Add New MCP Server.

  3. Name: youtube-transcript, Type: command.

  4. Command: uvx --from git+https://github.com/jkawamoto/mcp-youtube-transcript mcp-youtube-transcript

Goose

Please refer to this tutorial for detailed installation instructions: YouTube Transcript Extension.

LM Studio

To configure this server for LM Studio, click the button below.

Add MCP Server youtube-transcript to LM Studio

Using Docker

A Docker image for this server is available on Docker Hub. Please refer to the Docker Hub page for detailed usage instructions and documentation.

Available Tools

The server registers the following MCP tools for LLM agents:

Tool

Description

get_transcript

Retrieves the plain-text transcript for a YouTube video URL.

get_timed_transcript

Retrieves transcript segments with start times and durations.

get_available_languages

Lists all available transcript languages (manual & auto-generated).

get_video_info

Fetches video metadata such as title and channel information.

  • get_transcript

    • url (string, required): Full YouTube video URL.

    • lang (string, optional): Preferred language code (defaults to "en").

    • next_cursor (integer, optional): Cursor to fetch the next chunk for long videos.

  • get_timed_transcript

    • url (string, required): Full YouTube video URL.

    • lang (string, optional): Preferred language code (defaults to "en").

    • next_cursor (integer, optional): Cursor to fetch the next chunk.

  • get_available_languages

    • url (string, required): Full YouTube video URL.

  • get_video_info

    • url (string, required): Full YouTube video URL.

Advanced Configuration

Handling Long Videos (Pagination)

Long videos (e.g., lectures, conferences, podcast episodes) can quickly exceed LLM token context limits. By default, transcripts exceeding 50,000 characters are chunked. When a response is split, a next_cursor is provided so the LLM agent can autonomously query the rest.

To customize the chunk character limit, supply --response-limit. To disable pagination and fetch the entire transcript at once, set --response-limit to a negative value (e.g., -1):

{
  "mcpServers": {
    "youtube-transcript": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/jkawamoto/mcp-youtube-transcript",
        "mcp-youtube-transcript",
        "--response-limit",
        "15000"
      ]
    }
  }
}

Avoiding IP Bans (Proxy Setup)

YouTube aggressively blocks automated transcript requests from cloud providers and data center IPs. Using residential or rotating proxies ensures uninterrupted access.

1. Webshare Residential Proxy

Set credentials via environment variables or command-line flags:

{
  "mcpServers": {
    "youtube-transcript": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/jkawamoto/mcp-youtube-transcript",
        "mcp-youtube-transcript"
      ],
      "env": {
        "WEBSHARE_PROXY_USERNAME": "your_username",
        "WEBSHARE_PROXY_PASSWORD": "your_password"
      }
    }
  }
}

(CLI equivalents: --webshare-proxy-username and --webshare-proxy-password)

2. ScrapingAnt

If using ScrapingAnt, supply your API token:

{
  "mcpServers": {
    "youtube-transcript": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/jkawamoto/mcp-youtube-transcript",
        "mcp-youtube-transcript"
      ],
      "env": {
        "SCRAPINGANT_API_TOKEN": "your_api_token"
      }
    }
  }
}

(CLI equivalent: --scrapingant-api-token)

Accessing YouTube requires a paid ScrapingAnt Web Scraping API subscription. YouTube access requires residential proxies, which consume more ScrapingAnt credits than standard proxy requests. See the ScrapingAnt credit cost documentation for details.

3. Standard / Generic HTTP & HTTPS Proxies

Specify custom proxy endpoints via HTTP_PROXY / HTTPS_PROXY (or --http-proxy / --https-proxy):

{
  "mcpServers": {
    "youtube-transcript": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/jkawamoto/mcp-youtube-transcript",
        "mcp-youtube-transcript"
      ],
      "env": {
        "HTTPS_PROXY": "http://username:password@proxy.example.com:8080"
      }
    }
  }
}

For more details, please visit: Working around IP bans - YouTube Transcript API.

License

This application is licensed under the MIT License. See the LICENSE file for more details.

Available Tools

4 tools
get_available_languagesA

Retrieves the available languages for the video.

Inputs: url (required). Returns: result

get_timed_transcriptB

Retrieves the transcript of a YouTube video with timestamps.

Inputs: url (required), lang, next_cursor. Returns: Transcript of a YouTube video with timestamps.

get_transcriptC

Retrieves the transcript of a YouTube video.

Inputs: url (required), lang, next_cursor. Returns: Transcript of a YouTube video.

get_video_infoC

Retrieves the video information.

Inputs: url (required). Returns: Video information.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.5/5.0

    Scored across 4 tools

    Disambiguation4/5

    Each tool targets a distinct output type, though get_transcript and get_timed_transcript overlap heavily in purpose and are only separated by the presence of timestamps. get_available_languages and get_video_info are clearly distinct from transcript retrieval.

    Naming Consistency5/5

    All four tools follow a consistent get_noun snake_case pattern, making the set highly predictable and readable.

    Tool Count5/5

    Four tightly scoped tools is appropriate for a focused YouTube transcript server. Each tool earns its place without extraneous surface area.

    Completeness4/5

    Core needs—transcript text, timestamped transcript, language availability, and video metadata—are covered. Minor gaps like transcript translation or search within a transcript are not present but are not essential to the stated purpose.

    Maintenance

    ActivityActive
    ResponsivenessLimited data

    Related MCP Connectors

    Related MCP Servers