InternetSearch MCP

Enhanced web search and content extraction MCP server. Search the web, fetch pages, extract metadata, read RSS feeds, get YouTube transcripts, and more.

An evolution of mcp-server-fetch with significantly more capabilities.

Features

• Web Search — DuckDuckGo (default) and Brave Search API support
• Page Fetching — Extract clean text from any web page with pagination support
• Content Summarization — Auto-summarize pages using extractive summarization
• Parallel Fetching — Fetch multiple URLs at once
• RSS/Atom Feeds — Read and parse news feeds
• YouTube Transcripts — Extract video transcripts (requires yt-dlp)
• Metadata Extraction — OpenGraph, title, description, author, published date
• Link Extraction — Get all links from a page
• Search & Summarize — Search and get a summary of the top result in one call

Installation

npm install -g @nachoretro/internetsearch

Or use npx (no install):

npx @nachoretro/internetsearch

Configuration

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "internetsearch": {
      "command": "npx",
      "args": ["-y", "@nachoretro/internetsearch"],
      "env": {
        "BRAVE_API_KEY": "optional-brave-api-key-for-better-search"
      }
    }
  }
}

Environment Variables

| Variable | Description | |----------|-------------| | BRAVE_API_KEY | Brave Search API key for higher-quality search results |

Tools

web_search

Search the web using DuckDuckGo or Brave Search.

{
  "query": "latest AI developments 2025",
  "max_results": 10,
  "engine": "duckduckgo"
}

fetch_page

Fetch and extract content from a web page.

{
  "url": "https://example.com/article",
  "max_length": 5000,
  "start_index": 0,
  "extract_links": true,
  "raw": false
}

summarize_page

Fetch and summarize a page in one call.

{
  "url": "https://example.com/long-article",
  "sentences": 3
}

fetch_multiple

Fetch multiple URLs in parallel.

{
  "urls": [
    "https://example.com/page1",
    "https://example.com/page2",
    "https://example.com/page3"
  ],
  "max_length": 3000
}

read_rss

Read an RSS or Atom feed.

{
  "url": "https://news.ycombinator.com/rss",
  "max_items": 10
}

youtube_transcript

Extract transcript from a YouTube video (requires yt-dlp installed).

{
  "video_id": "dQw4w9WgXcQ"
}

Or with full URL:

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

extract_metadata

Extract metadata from a URL.

{
  "url": "https://example.com/article"
}

Returns: title, description, author, published date, OpenGraph image, word count.

search_and_summarize

Search and summarize the top result in one call.

{
  "query": "climate change latest research",
  "engine": "duckduckgo"
}

Comparison with mcp-server-fetch

| Feature | mcp-server-fetch | internetsearch | |---------|------------------|----------------| | Basic page fetch | ✅ | ✅ | | Content extraction | ✅ | ✅ (improved) | | Pagination | ✅ | ✅ | | Web search | ❌ | ✅ DuckDuckGo + Brave | | Summarization | ❌ | ✅ Auto-summarize | | Parallel fetch | ❌ | ✅ Multiple URLs | | RSS feeds | ❌ | ✅ RSS + Atom | | YouTube transcripts | ❌ | ✅ Requires yt-dlp | | Metadata extraction | ❌ | ✅ OpenGraph, etc. | | Search & summarize | ❌ | ✅ Combined operation | | Link extraction | ❌ | ✅ All page links | | HTML parsing | readabilipy | cheerio (lighter) |

Prerequisites

Optional (for enhanced features)

1. ddgr — For DuckDuckGo search via command line:

   pip install ddgr

2. yt-dlp — For YouTube transcript extraction:

   pip install yt-dlp

3. Brave Search API Key — For higher quality search results:

   • Get API key at brave.com/search/api
   • Set BRAVE_API_KEY environment variable

Usage Examples

Research Workflow

User: "What's the latest on quantum computing?"

→ web_search: "quantum computing breakthroughs 2025"
→ fetch_page: top result URL
→ summarize_page: for quick overview

Content Monitoring

User: "Check the latest from Hacker News"

→ read_rss: "https://news.ycombinator.com/rss"
→ fetch_multiple: top 5 story URLs

Video Analysis

User: "Summarize this YouTube video"

→ youtube_transcript: video_id
→ (Claude summarizes the transcript)

Development

git clone https://github.com/hubinoretros/internetsearch.git
cd internetsearch
npm install
npm run build
npm start

License

MIT