Markdown

You can run a dedicated Markdown MCP Server that uses Playwright to extract clean, readable markdown content from web pages. It identifies the main article content, removes navigation and boilerplate, and preserves formatting like headings, lists, code blocks, tables, and images or links when enabled. This lets you fetch consistent, portable Markdown from any URL for downstream workflows or AI tools.

How to use

You will run a local MCP server and connect your MCP client to it. When you send a request to the server, specify the target URL you want converted to markdown. You can tailor the extraction with a few options: includeImages to add image references, includeLinks to preserve hyperlinks, waitForSelector to pause until dynamic content is loaded, and timeout to bound how long the page may take.

How to install

Prerequisites you need ahead of time:

• Node.js 18 or higher must be installed on your machine.

• Git to clone the project.

• npm to install dependencies.

Install Playwright browsers (Chromium) as part of the setup.

Follow these concrete steps to set up the server locally:

Install and run steps (concrete commands)

# Step 1: Clone the MCP server repository
git clone <repository-url>
cd markdown-mcp

# Step 2: Install dependencies
npm install

# Step 3: Install Playwright browsers
npx playwright install chromium

# Step 4: Make the main script executable (optional but recommended)
chmod +x markdown-mcp.js

# Step 5: Run the MCP server (stdin/stdout configuration)
node markdown-mcp.js

Additional configuration and usage details

This MCP server provides a single tool named get_page_markdown. You can pass options such as url, includeImages, includeLinks, waitForSelector, and timeout to customize the extraction. Use the tool through your MCP client to obtain clean markdown from any web page.

Configuration and security notes

• Each request uses a fresh browser context for isolation and security.
• The server runs in headless mode to minimize exposure.
• No cookies or persistent data are stored between requests.
• Timeouts limit network requests to protect against unresponsive pages.

Testing and troubleshooting

Test by starting the server and sending a simple extraction request to verify the output. If you encounter issues, check that Node.js and npm are installed, ensure the script has execute permissions, and run the server manually to observe errors.

Notes on client integrations

You can use the same Markdown MCP Server with multiple MCP clients that support the Model Context Protocol. The server is designed to handle concurrent requests efficiently and produce consistent markdown results across clients.

Tools exposed by the server

• get_page_markdown: Extracts clean markdown content from a given URL, with options to include images and links, and to wait for dynamic content to load.

Available tools

get_page_markdown

Tool that extracts main content from a URL and returns it as Markdown, with options to include images and links, wait for a selector, and set a timeout.