# @fabriqa.ai/youtube-transcript-mcp [Health: Active]

**Category:** 🌐 Social Media  
**Repository:** https://github.com/hancengiz/youtube-transcript-mcp  
**GitHub Stars:** 15  
**npm Downloads (last month):** 376  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/fabriqa-ai-youtube-transcript-mcp

## Description
MCP server for retrieving YouTube video transcripts for Claude Code

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "youtube-transcript-mcp": {
    "command": "npx",
    "args": ["-y","@fabriqa.ai/youtube-transcript-mcp@latest"]
  }
}
```

## Documentation

## What @fabriqa.ai/youtube-transcript-mcp does

@fabriqa.ai/youtube-transcript-mcp exposes YouTube transcript retrieval through the Model Context Protocol. It is intended for workflows where an AI assistant needs the text of a captioned video rather than video metadata or a downloaded media file. The returned transcript can then be summarized, searched, quoted, or analyzed by the connected client.

The package accepts common YouTube links, including standard watch URLs, shortened `youtu.be` links, mobile URLs, and direct 11-character video IDs. It can request a particular language when that language is available and can return text with or without timestamps.

## Tools and capabilities

The server provides two tools:

- `get-transcript` retrieves a video's transcript. Its required `url` parameter accepts a YouTube URL or video ID. The optional `lang` parameter selects a language such as `en`, `es`, or `fr`; without it, the video's default language is used. The `include_timestamps` option controls whether timestamps are included and defaults to `true`.
- `get-transcript-languages` checks which transcript languages are available for a specified video. It accepts the same URL or video ID style as the transcript tool.

The server focuses on transcript content, which avoids loading unrelated video metadata. Timestamp-free output can reduce the amount of context consumed when processing long videos.

## Setup and configuration

Install the package globally with npm:

```bash
npm install -g @fabriqa.ai/youtube-transcript-mcp
```

For a source installation, clone the repository, enter its directory, and run `npm install`.

Claude Code can register the package with `npx`. A user-scoped configuration makes it available across projects:

```bash
claude mcp add --scope user youtube-transcript npx @fabriqa.ai/youtube-transcript-mcp@latest
```

Use `--scope local` for a project-specific configuration instead. The package also includes an update script that can add the server to Claude's configuration:

```bash
npx @fabriqa.ai/youtube-transcript-mcp/update-config.js
```

For Claude Desktop, add an `mcpServers` entry to the platform-specific Claude Desktop configuration file:

```json
{
  "mcpServers": {
    "youtube-transcript": {
      "command": "npx",
      "args": ["@fabriqa.ai/youtube-transcript-mcp@latest"]
    }
  }
}
```

Restart the client after configuration so the tools become available.

## Limitations and notes

Transcript retrieval depends on captions being available for the selected YouTube video. Invalid URLs and videos without transcripts are handled with errors rather than producing transcript content. Requested languages also need to be available for the video. The server retrieves transcript text; it is not described as a video downloader or a general YouTube metadata client.

For long videos, omitting timestamps can produce a smaller result. The package's example notes that a 60-minute transcript may be about 19,000 tokens without timestamps compared with about 30,000 with them. Claude Code users can also place transcript analysis in a dedicated sub-agent workflow to keep large transcript content out of the main conversation.

_Full upstream README: https://allmcps.com/mcp/fabriqa-ai-youtube-transcript-mcp/readme_