YouTube Transcript MCP Server: Setup Guide, Use Cases & Limitations

What Is YouTube Transcript MCP Server?

YouTube Transcript MCP Server is an MCP server that fetches and searches transcripts from YouTube videos directly within Claude. It is maintained by kimtaeyoon83 and published as an npm package.

Quick Facts

What You Can Do With It

• Transcript extraction: Pull the full transcript from any YouTube video that has captions available, using a video URL or ID.
• Video summarization: Feed a transcript directly into Claude to generate concise summaries of long-form video content.
• Content research: Retrieve transcripts from multiple videos to compare talking points, arguments, or coverage of a topic.
• Quote finding: Search transcript text to locate specific statements, timestamps, or phrasing within a video.

How to Set It Up

Prerequisites: Node.js installed on your machine.

Step 1 — Verify the package runs locally:

npx mcp-youtube-transcript
**Step 2 — Add the server to your Claude configuration.**

For **Claude Desktop**, edit `claude_desktop_config.json` (typically at `~/Library/Application Support/Claude/claude_desktop_config.json` on macOS):

```json
{
  "mcpServers": {
    "youtube-transcript": {
      "command": "npx",
      "args": ["mcp-youtube-transcript"]
    }
  }
}

For Claude Code, add the server via the CLI or edit .claude/mcp.json in your project root:

{
  "mcpServers": {
    "youtube-transcript": {
      "command": "npx",
      "args": ["mcp-youtube-transcript"]
    }
  }
}

Step 3 — Restart Claude Desktop or reload your Claude Code session. The server should appear as an available tool. Test it by asking Claude to fetch the transcript of a YouTube URL.

Works Best With

This server is documented as compatible with Claude Desktop and Claude Code. Both clients support the MCP tool-calling protocol required to pass YouTube URLs to the server and return transcript text. Claude Code is particularly useful when combining transcript retrieval with file I/O — for example, saving transcripts to disk or piping content into other MCP tools in a multi-server workflow.

Known Limitations

• Captions required: The server can only extract transcripts where YouTube captions exist — either auto-generated or manually uploaded. Videos with no captions will return an error or empty result.
• Auto-generated caption quality: Auto-generated transcripts often lack punctuation and may contain transcription errors, which affects downstream summarization or quote accuracy.
• No YouTube authentication: The server accesses public caption data only. Private videos, members-only content, and age-restricted videos that require login are not accessible.
• Rate limiting: YouTube’s internal caption endpoints can throttle repeated requests. Batch processing many videos in quick succession may result in failures or incomplete responses.
• Language selection: If a video has multiple caption tracks, verify the server’s behavior for selecting the default language — consult the GitHub repository for current handling.

Community Resources

• GitHub Issues: https://github.com/kimtaeyoon83/mcp-server-youtube-transcript/issues — report bugs or check for known problems before opening a new issue.
• GitHub Repository: https://github.com/kimtaeyoon83/mcp-server-youtube-transcript — source code, README, and contribution guidelines.
• TrendScoped MCP Directory: https://trendscoped.com/mcp-mcp-youtube-transcript — this listing page for community context and updates.