Toon Fetch

Toto Fetch MCP Server lets you retrieve web page content, convert it to Markdown, and optionally run a local language model to further process the text before returning a structured TOON payload for your agents. It streamlines getting clean, readable article content from any URL and, if you want, adds AI-assisted insights right where you need them.

How to use

You use this server by configuring an MCP client to connect to the provided fetch_url tool. When you call fetch_url with a URL, you can optionally pass a prompt to guide the local AI processing. If you omit a prompt, you simply receive the raw Markdown content of the article. The output is formatted as a TOON object, optimized for agent consumption, including the page title, the Markdown content, and relevant metadata such as the source URL and an excerpt.

How to install

Prerequisites: Node.js and npm must be available on your system.

# Install dependencies
npm install

# Build the project
npm run build

Configuration and server setup

To run this as an MCP server, you expose it to your MCP client using the following configuration snippet. This tells your MCP client how to start the server as a stdio process using npx.

{
  "mcpServers": {
    "toon-fetch": {
      "command": "npx",
      "args": ["toon-fetch"]
    }
  }
}

Available tools

fetch_url

Fetches HTML from a URL, extracts the main content, converts it to Markdown, and optionally processes it with a local LLM. Returns data in TOON format.