home / mcp / superfetch mcp server
Provides clean, AI-ready Markdown by fetching public URLs and extracting readable content with safety-first defaults.
Configuration
View docs{
"mcpServers": {
"j0hanz-super-fetch-mcp-server": {
"url": "http://127.0.0.1:3000/mcp",
"headers": {
"HOST": "127.0.0.1",
"PORT": "3000",
"CACHE_TTL": "3600",
"LOG_LEVEL": "info",
"USER_AGENT": "superFetch-MCP/2.0",
"ALLOW_REMOTE": "false",
"CACHE_ENABLED": "true"
}
}
}
}You can run a dedicated MCP server called superFetch that fetches web pages, extracts readable content with Mozilla Readability, and returns clean, AI-friendly Markdown. It’s designed for AI workflows like summarization, context retrieval, and knowledge-base ingestion, with safety defaults and flexible client integrations.
Run the server in stdio mode for direct MCP client integration, or run in HTTP mode for remote access with authentication. Use the MCP client to connect, then call the fetch-url tool to convert any public webpage into clean Markdown with a metadata footer. The output includes a structured content object with the original URL, the resolved URL, an optional title, and the Markdown content.
- In stdio mode, connect to the server via your MCP client using the provided npm/npx-based command, then issue the fetch-url operation to retrieve Markdown from a URL.
Prerequisites: you need Node.js (version 20 or newer) and npm. You will run the MCP server via npm/npx as shown.
{
"mcpServers": {
"superFetch": {
"command": "npx",
"args": ["-y", "@j0hanz/superfetch@latest", "--stdio"]
}
}
}Step-by-step minimal flow to get started in stdio mode: install the package, then start in stdio mode using the MCP client. If you prefer HTTP mode, run the HTTP server command as shown in the HTTP configuration example.
The server offers two main modes: - stdio mode: ideal for MCP clients like VS Code, Claude Desktop, Cursor, and similar tools. Use the provided stdio command to run the MCP server locally and connect directly. - HTTP mode: suitable for self-hosted setups with authentication, rate limiting, and host/origin validation. This mode exposes several endpoints under /mcp and /mcp/downloads and requires an API token or OAuth configuration.
Core server settings you may configure via environment variables include host, port, caching behavior, timeouts, and safety features. See the list of core settings below for the exact variable names and defaults.
SSRF protection blocks private IP ranges, cloud metadata endpoints, and internal host suffixes. URL validation restricts to http/https URLs without embedded credentials and limits URL length. In HTTP mode, Host and Origin headers are validated against an allow-list. Rate limiting is applied to relevant endpoints to prevent abuse.
If you encounter issues on Windows with the stdio workflow, try running the exact command shown in the configuration blocks, ensuring your shell environment has Node and npm available.
Fetches a webpage and converts it to clean Markdown with a metadata footer for HTML. It preserves raw Markdown when applicable and rewrites raw content URLs for direct access.