markudown-mcp

MCP server for MarkUDown / Scrape Technology — gives AI agents (Claude, Cursor, Windsurf, n8n, LangGraph, etc.) access to web scraping, Google search, data extraction, and more.

No infrastructure needed. Just an API key.

How it works

AI Agent (Claude Desktop, Cursor, etc.)
        │  MCP protocol (stdio)
        ▼
  markudown-mcp  (this package)
        │  HTTPS + polling
        ▼
  api.scrapetechnology.com
        │  BullMQ jobs
        ▼
  MarkUDown Engine workers

Quick start

MARKUDOWN_API_KEY=your-key npx markudown-mcp

Get your API key at scrapetechnology.com.

Claude Desktop

Add to claude_desktop_config.json:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "markudown": {
      "command": "npx",
      "args": ["markudown-mcp"],
      "env": {
        "MARKUDOWN_API_KEY": "your-api-key-here"
      }
    }
  }
}

Restart Claude Desktop after saving.

Cursor / Windsurf

Add to your editor's MCP config:

{
  "markudown": {
    "command": "npx",
    "args": ["markudown-mcp"],
    "env": {
      "MARKUDOWN_API_KEY": "your-api-key-here"
    }
  }
}

Environment variables

Available tools

scrape

Scrape a single URL and return its content as clean Markdown.

Example prompt: "Scrape https://example.com and summarize the main content."

map

Discover all URLs on a website via sitemap parsing and link crawling.

Example prompt: "Map all URLs on docs.example.com."

crawl

Recursively crawl a website, following internal links. Returns Markdown for each page.

Example prompt: "Crawl https://docs.example.com up to depth 3 and find all mentions of 'authentication'."

search

Run a Google search and optionally scrape the full content of each result page.

Example prompt: "Search for 'best web scraping libraries 2025' and summarize the top 5 results."

extract

Scrape a URL and extract structured data using a natural language prompt.

Example prompt: "Extract all job listings from https://jobs.example.com including title, location, and salary."

batch_scrape

Scrape multiple URLs in parallel. Returns Markdown for each.

Example prompt: "Scrape these 5 competitor pricing pages and compare their plans."

screenshot

Take a full-page screenshot of a URL.

Example prompt: "Take a screenshot of https://example.com."

deep_research

Scrape multiple URLs and synthesize their content into a comprehensive research report using an LLM.

Example prompt: "Research the pricing strategies of these 3 SaaS companies and write a comparative analysis."

agent

Launch an AI agent that autonomously navigates a website to answer a question or complete a task.

Example prompt: "Go to https://shop.example.com and find the return policy."

change_detection

Detect whether a URL's content has changed since it was last checked.

Example prompt: "Check if the pricing page at https://example.com/pricing has changed."

Self-hosted deployment

If you run your own MarkUDown Engine, set MARKUDOWN_API_URL to point to your instance:

{
  "mcpServers": {
    "markudown": {
      "command": "npx",
      "args": ["markudown-mcp"],
      "env": {
        "MARKUDOWN_API_KEY": "your-api-key",
        "MARKUDOWN_API_URL": "https://your-engine.example.com"
      }
    }
  }
}

For direct BullMQ access (no API key needed), use the self-hosted MCP server in MarkUDown-Engine/services/mcp/.

Support & Community

License

MIT