Defuddle Fetch

The Defuddle Fetch MCP Server lets you retrieve web content and convert it into clean, readable markdown (or raw HTML) for use by large language models. It automatically strips clutter from pages, preserves key metadata like the title and author, and supports paginated outputs for long content. This makes it easy to feed rich web content into your AI workflows with consistent formatting.

How to use

Use the Defuddle Fetch MCP Server through an MCP client to fetch content from any web URL. You provide a URL and optional parameters to control how much content is returned and how it is formatted. The server returns the cleaned content along with metadata such as title, author, publication date, word count, and the domain of the source. You can request the content as markdown for easy inclusion in prompts or as raw HTML if you need to preserve formatting for further processing.

How to install

Prerequisites: you need Node.js and npm installed on your system.

Install and run the MCP server via a local development command. The following configuration shows two common ways to integrate Defuddle Fetch with your environment.

{
  "mcpServers": {
    "defuddle-fetch": {
      "command": "npx",
      "args": [
        "-y",
        "defuddle-fetch-mcp-server"
      ]
    }
  }
}

Alternatively, after you clone the project and build it, you can start the local server with Node by pointing to the built file. The configuration example shows how Claude Desktop would launch the server from a dist/index.js file.

{
  "mcpServers": {
    "defuddle-fetch": {
      "command": "node",
      "args": [
        "/path/to/clone/defuddle-fetch-mcp-server/dist/index.js"
      ]
    }
  }
}

Additional notes

Tools exposed by this MCP server include a fetch endpoint that retrieves a URL and returns cleaned content with optional metadata. The fetch tool accepts the following inputs and outputs: input url (string, required) to fetch; max_length (number, optional) to cap the character length; start_index (number, optional) to begin reading from a character index; raw (boolean, optional) to return unformatted content. The response includes title, author, publication date, word count, domain, and processing time.

If you need to customize behavior for pagination or output formatting, pass start_index and max_length to control chunked reading and choose between markdown or raw HTML output via the raw flag.

Available tools

fetch

Fetches a URL from the internet and extracts its contents as clean, markdown text using Defuddle. Returns content with metadata including title, author, publication date, word count, domain, and processing time. Input parameters: url (string, required), max_length (number, optional, defaults to 5000), start_index (number, optional, defaults to 0), raw (boolean, optional, defaults to false). Output includes processed content and metadata.