yt-dlp-mcp

An MCP server that gives AI assistants the ability to retrieve transcripts, comments, and metadata from online videos. Works with any site supported by yt-dlp — YouTube, Vimeo, Twitch, and thousands more.

Features

• Transcripts — Download subtitles in plain text or structured JSON, with optional timestamps and time-range filtering for long videos
• Comments — Fetch top comments sorted by popularity, including author info, likes, and pinned status
• Subtitle tracks — List all available subtitle languages and formats for a video
• Video metadata — Title, channel, duration, view count, and more, included with every response
• Video insights workflow — Built-in MCP resource with a step-by-step guide for extracting key insights from videos

Prerequisites

• Node.js (v18+)
• yt-dlp installed and available on your PATH

Install yt-dlp:

# pip
pip install yt-dlp

# brew
brew install yt-dlp

# or download from https://github.com/yt-dlp/yt-dlp/releases

Installation

Claude Desktop / Claude Code

Add to your MCP server configuration:

{
  "mcpServers": {
    "yt-dlp": {
      "command": "npx",
      "args": ["-y", "@akp-tools/yt-dlp-mcp"]
    }
  }
}

Global install

npm install -g @akp-tools/yt-dlp-mcp

Then run directly:

yt-dlp-mcp

Tools

list_subtitles

Lists available subtitle tracks for a video. Returns video metadata and all available subtitle languages/formats.

url: "https://www.youtube.com/watch?v=dQw4w9WgXcQ"

get_transcript

Downloads and returns the transcript of a video.

| Parameter | Required | Default | Description | |-----------|----------|---------|-------------| | url | Yes | — | Video URL | | language | No | en | Subtitle language code | | include_timestamps | No | false | Add [HH:MM:SS] timestamps | | output_format | No | text | text or json | | start_time | No | — | Start of time range (HH:MM:SS or MM:SS) | | end_time | No | — | End of time range (HH:MM:SS or MM:SS) |

For long videos, use start_time and end_time to fetch specific segments instead of the full transcript.

get_comments

Fetches top comments sorted by popularity.

| Parameter | Required | Default | Description | |-----------|----------|---------|-------------| | url | Yes | — | Video URL | | max_comments | No | 20 | Number of comments to fetch (1–500) |

get_resource

Retrieves built-in workflow guides. Useful as a fallback for MCP clients that don't support the resources protocol.

| Parameter | Required | Default | Description | |-----------|----------|---------|-------------| | resource | No | — | Resource name (omit to list all available) |

Resources

| URI | Description | |-----|-------------| | yt-dlp://instructions/video-insights | Step-by-step workflow for extracting key insights from a video using transcripts and comments |

Development

# Install dependencies
pnpm install

# Build
pnpm build

# Watch mode
pnpm dev

# Run tests
pnpm test

License

MIT