webread-mcp 1.0.0

MCP server for reading webpages and web search

WebRead MCP

A Model Context Protocol (MCP) server that enables AI assistants to read webpages and perform web searches.

Tools

read_webpage

Fetches a webpage via GET request and returns its text content.

Parameters:

• url (required) - The URL to fetch
• max_chars (optional, default: 500) - Maximum characters to return
• offset (optional, default: 0) - Starting character position for chunked reading
• raw_html (optional, default: false) - Return raw HTML instead of extracted text

Example:

{
  "url": "https://example.com",
  "max_chars": 1000,
  "offset": 0
}

web_search

Performs a web search using DuckDuckGo and returns titles with links.

Parameters:

• query (required) - The search query
• page (optional, default: 1) - Page number (1-based)

Example:

{
  "query": "python tutorial",
  "page": 1
}

Installation

From PyPI

pip install webread-mcp

From Source

git clone <repository-url>
cd webreadmcp
pip install -e .

Configuration

Add the server to your MCP client configuration:

If installed via pip

{
  "mcpServers": {
    "webread": {
      "command": "webread-mcp"
    }
  }
}

If running from source

{
  "mcpServers": {
    "webread": {
      "command": "python",
      "args": ["/absolute/path/to/webreadmcp/server.py"]
    }
  }
}

Running Manually

webread-mcp

Or from source:

python server.py

The server communicates via stdio and will wait for MCP protocol messages.

Dependencies

• mcp - Model Context Protocol SDK
• httpx - HTTP client
• beautifulsoup4 - HTML parsing
• duckduckgo_search - Web search