Getting Started

Automatic Installation

The quickest way to get started with ModelFetch is to use our create-modelfetch CLI. It sets up everything you need in seconds.

npx -y create-modelfetch@latest

The CLI will guide you through:

1. Choosing a runtime - Node.js, Bun, Deno, AWS Lambda, Vercel, Cloudflare, or Netlify
2. Selecting a language - TypeScript or JavaScript
3. Picking a package manager - npm, pnpm, bun, or yarn

Manual Installation

If you prefer to set up manually, install the ModelFetch package for your runtime:

npm install @modelfetch/node

Then, create the following files:

Server File

Create a server file to define your McpServer using the official MCP TypeScript SDK:

import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
import { z } from "zod";

const server = new McpServer({
  title: "My MCP Server",
  name: "my-mcp-server",
  version: "1.0.0",
});

server.registerTool(
  "roll_dice",
  {
    title: "Roll Dice",
    description: "Rolls an N-sided dice",
    inputSchema: { sides: z.number().int().min(2) },
  },
  ({ sides }) => ({
    content: [
      {
        type: "text",
        text: `🎲 You rolled a ${1 + Math.floor(Math.random() * sides)}!`,
      },
    ],
  }),
);

export default server;

Entry Point File

Create an entry point file to run your McpServer using the ModelFetch package for your runtime:

import handle from "@modelfetch/node"; // Choose your runtime
import server from "./server"; // Import your server

handle(server); // Let ModelFetch handle all runtime-specific details

Further Reading