MCP 服务器

deep-search-mcp

Performs comprehensive web searches by combining Google search with advanced content extraction using Mozilla's Readability algorithm.

README

Deep Search MCP Server

An MCP (Model Context Protocol) server that performs comprehensive web searches by combining Google search with advanced content extraction using Mozilla's Readability algorithm.

Features

Advanced Content Extraction - Uses Mozilla's Readability algorithm (same as Firefox Reader View) for clean article extraction
Multiple Search Types - Web search, news search, and image search
Domain Filtering - Include or exclude specific domains from results
Retry Logic - Automatic retries with exponential backoff for reliability
Controlled Concurrency - Fetches pages in batches to avoid overwhelming servers
Full Content - Returns complete page content, not just snippets

Prerequisites

Get a Serper API Key

This MCP server uses Serper.dev for Google search results.

Go to https://serper.dev
Sign up for a free account (2,500 free searches)
Copy your API key from the dashboard

Installation

Using npx (Recommended)

No installation needed - just configure your MCP client:

{
  "mcpServers": {
    "deep-search": {
      "command": "npx",
      "args": ["-y", "@thejusdutt/deep-search-mcp"],
      "env": {
        "SERPER_API_KEY": "your-serper-api-key-here"
      }
    }
  }
}

Global Installation

npm install -g @thejusdutt/deep-search-mcp

Then configure:

{
  "mcpServers": {
    "deep-search": {
      "command": "deep-search-mcp",
      "env": {
        "SERPER_API_KEY": "your-serper-api-key-here"
      }
    }
  }
}

Tools

`deep_search`

Comprehensive web search with full content extraction.

Parameters:

Parameter	Type	Default	Description
`query`	string	required	The search query
`num_results`	number	10	Number of results (1-10)
`max_content_per_page`	number	50000	Max characters per page (5000-100000)
`search_type`	string	"web"	Search type: "web", "news", or "images"
`include_domains`	string	-	Comma-separated domains to include
`exclude_domains`	string	-	Comma-separated domains to exclude

Examples:

// Basic web search
deep_search({ query: "React best practices 2025" })

// News search
deep_search({ query: "AI announcements", search_type: "news" })

// Image search - returns image URLs and source pages
deep_search({ query: "cute cats", search_type: "images" })

// Search specific sites only
deep_search({ 
  query: "TypeScript tips",
  include_domains: "github.com,dev.to"
})

// Exclude certain sites
deep_search({
  query: "web development trends",
  exclude_domains: "pinterest.com,facebook.com"
})

`deep_search_news`

Optimized for news article search.

Parameters:

Parameter	Type	Default	Description
`query`	string	required	The news topic to search
`num_results`	number	10	Number of articles (1-10)
`max_content_per_page`	number	30000	Max characters per article

Example:

deep_search_news({ query: "OpenAI latest updates" })

Configuration for Different MCP Clients

Kiro / Claude Desktop

Add to ~/.kiro/settings/mcp.json or claude_desktop_config.json:

{
  "mcpServers": {
    "deep-search": {
      "command": "npx",
      "args": ["-y", "deep-search-mcp"],
      "env": {
        "SERPER_API_KEY": "your-api-key"
      }
    }
  }
}

VS Code with Continue

Add to your Continue config:

{
  "mcpServers": [
    {
      "name": "deep-search",
      "command": "npx",
      "args": ["-y", "deep-search-mcp"],
      "env": {
        "SERPER_API_KEY": "your-api-key"
      }
    }
  ]
}