urltomarkdown-mcp

An MCP server that converts URLs and raw HTML to clean Markdown. Built on top of urltomarkdown.

Give any MCP-compatible AI assistant the ability to read web pages — just point it at a URL and get structured Markdown back, ready for summarization, analysis, or ingestion into your workflow.

Tools

convert_url_to_markdown

Fetches a URL and converts the web page to clean Markdown.

convert_html_to_markdown

Converts raw HTML to clean Markdown (no network request needed).

Use cases

• Research — Have your AI assistant read and summarize articles, documentation, or blog posts

• Data extraction — Pull structured content from web pages for analysis

• Documentation ingestion — Convert API docs or reference pages into Markdown for context

• Content migration — Bulk convert web content to Markdown format

• RAG pipelines — Feed clean web content into retrieval-augmented generation systems

Installation

Using npx (no install needed)

{
  "mcpServers": {
    "urltomarkdown": {
      "command": "npx",
      "args": ["-y", "urltomarkdown-mcp"]
    }
  }
}

Global install

npm install -g urltomarkdown-mcp

Then add to your MCP config:

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

From source

git clone https://github.com/SapienEx-AI/urltomarkdown-mcp.git
cd urltomarkdown-mcp
npm install

{
  "mcpServers": {
    "urltomarkdown": {
      "command": "node",
      "args": ["/path/to/urltomarkdown-mcp/src/index.js"]
    }
  }
}

Where to put the config

How it works

This server wraps the urltomarkdown library and exposes it over the Model Context Protocol via stdio transport. Under the hood it uses:

• Turndown for HTML-to-Markdown conversion

• Mozilla's Readability for content extraction and noise removal

• JSDOM for DOM parsing

The Readability pass strips navigation, sidebars, ads, and other non-content elements before conversion, producing clean output suitable for LLM consumption.

Testing

npm test

License

MIT

SapienEx — AI strategy and consulting