YouTube & Web Search MCP Server

A Model Context Protocol (MCP) server that provides tools to extract YouTube video IDs, fetch transcripts, and search the web using Tavily API.

Installation

Using uv (recommended)

# Install all dependencies from pyproject.toml
uv sync

# Or install specific packages (if starting fresh)
uv add "mcp[cli]" youtube-transcript-api tavily-python

Usage

Run server

uv run youtube_mcp_server.py

Development Mode

Test the server with MCP Inspector:

uv run mcp dev youtube_mcp_server.py

Claude Desktop Integration

Install in Claude Desktop:

uv run mcp install youtube_mcp_server.py --name "YouTube Tools"

Direct Execution

python youtube_mcp_server.py

Environment Variables

For web search functionality, you need to set your Tavily API key:

export TAVILY_API_KEY="your_tavily_api_key_here"

Get your free API key at: https://tavily.com

Available Tools

1. get_youtube_video_id(url): Extract video ID from YouTube URLs
2. get_youtube_transcript(video_url_or_id, language="en"): Get full transcript with timestamps
3. search_web(query, max_results=5, include_domains=None, exclude_domains=None): Search the web using Tavily API

Example Usage

YouTube Tools

The tools can extract video IDs from various YouTube URL formats:

• https://www.youtube.com/watch?v=dQw4w9WgXcQ
• https://youtu.be/dQw4w9WgXcQ
• dQw4w9WgXcQ (direct video ID)

Transcripts are returned with timestamps in [MM:SS] text format.

Web Search

The web search tool provides comprehensive search results with quick answers:

• Basic search: search_web("artificial intelligence trends 2024")
• Limited results: search_web("python tutorials", max_results=3)
• Domain filtering: search_web("machine learning", include_domains=["wikipedia.org", "arxiv.org"])

Results include titles, URLs, content snippets, and often a quick answer summary.