HTTP Fetch

You run a lightweight MCP server that fetches web pages and converts them to Markdown, enabling easy integration with MCP clients to retrieve clean, readable content from any URL. This server supports both web-friendly SSE communication and local stdio processes for flexible deployment patterns.

How to use

You connect to the server using an MCP client and invoke the fetch_url tool to retrieve any web page as Markdown. Use SSE for web-based deployments, APIs, or Kubernetes setups, or use stdio for desktop clients and sidecar-style workflows. The server exposes one MCP tool named fetch_url which accepts a URL and returns the page converted to Markdown.

How to install

Prerequisites: Python 3.8+ and Git must be available on your system. You also need a working internet connection to install dependencies.

Step 1: Clone the repository.

Step 2: Install dependencies.

Step 3: Run the server. Choose SSE for web/Kubernetes deployments or stdio for desktop/sidecar use.

# SSE protocol (default - web/Kubernetes deployments)
python app/server.py

# Custom host/port for Kubernetes
python app/server.py --protocol sse --host 0.0.0.0 --port 8000

# Stdio protocol (for desktop clients/sidecars)
python app/server.py --protocol stdio

Additional setup and options

Docker is supported for containerized deployments. Build and run the image to get started quickly.

The server also supports a convenience script for quick startups.

Configuration, security, and notes

Configuration includes a 30-second timeout for HTTP requests and support for HTML/XML content. Markdown output preserves links and images from the source content.

Security considerations include validating URLs, running in a non-root container, and avoiding execution of JavaScript content. The server is designed to prevent hanging by applying reasonable timeout limits.

For testing, you can run the included test suite and a demo to verify HTML-to-Markdown conversion.

Available tools

fetch_url

Fetch a web page and convert it to Markdown format