MCP NPX Fetch

What is MCP NPX Fetch?

MCP NPX Fetch is an MCP server that fetches web content and transforms it into formats such as HTML, JSON, Markdown, or plain text, designed for integration with MCP clients like Claude Desktop.

How to use MCP NPX Fetch?

Install globally via npm install -g @tokenizin/mcp-npx-fetch or run directly with npx @tokenizin/mcp-npx-fetch. Add a configuration entry to your Claude Desktop config to use it as an MCP server, then call any of its four tools (fetch_html, fetch_json, fetch_txt, fetch_markdown) from your MCP client.

Key features of MCP NPX Fetch

• Supports fetching content in HTML, JSON, text, and Markdown formats
• Custom request headers for authentication and personalization
• Automatic conversion between content formats
• Seamless integration with Claude Desktop and other MCP clients
• Full TypeScript type definitions for type safety
• Built on JSDOM, Turndown, Zod, and the MCP SDK

Use cases of MCP NPX Fetch

• Extract clean text from any webpage for LLM consumption
• Fetch and parse JSON APIs directly into structured data
• Convert web pages to Markdown for documentation or note-taking
• Add authentication headers to access private content

FAQ from MCP NPX Fetch

What is the difference between the four tools?

fetch_html returns raw HTML, fetch_json parses JSON, fetch_txt strips HTML to plain text, and fetch_markdown converts HTML to Markdown.

What runtime dependencies are required?

Node.js and npm; the server uses JSDOM, Turndown, the MCP SDK, and Zod under the hood.

How do I pass custom headers?

All four tools accept an optional headers object where you can provide key-value pairs for request headers.

Does MCP NPX Fetch support authentication?

Yes, you can include authentication tokens or API keys via the custom headers parameter.

Where does the fetched data reside?

The server returns the fetched content directly in the tool response; no data is stored or cached by the server.