ModelFetch is a delightful TypeScript/JavaScript SDK for building and deploying MCP servers anywhere TypeScript/JavaScript runs.
- Multi-Runtime: Write once, run anywhere: Node.js, Next.js, Bun, Deno, AWS Lambda, Vercel, Cloudflare, etc.
- Official SDK: Built on top of the official MCP TypeScript SDK to avoid lock-in, guarantee long-term support, and ensure up-to-date implementation
- Live Reload: Development server with automatic reloading
- MCP Inspector: Built-in integration for testing and debugging
- Modular Design: Platform-specific package for optimal performance
Get started in seconds using our create-modelfetch
CLI:
npx -y create-modelfetch@latest
ModelFetch works with any McpServer
instance from the official MCP TypeScript SDK. Here's all it takes:
Create your McpServer
with 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;
Run it anywhere with ModelFetch's handle()
function
import handle from "@modelfetch/node"; // Choose your runtime
import server from "./server"; // Import your server
handle(server); // That's it — ModelFetch handles all runtime-specific details
import handle from "@modelfetch/next"; // Choose your runtime
import server from "./server"; // Import your server
const handler = handle(server); // That's it — ModelFetch handles all runtime-specific details
export const GET = handler;
export const POST = handler;
export const DELETE = handler;
import handle from "@modelfetch/bun"; // Choose your runtime
import server from "./server.ts"; // Import your server
handle(server); // That's it — ModelFetch handles all runtime-specific details
import handle from "@modelfetch/deno"; // Choose your runtime
import server from "./server.ts"; // Import your server
handle(server); // That's it — ModelFetch handles all runtime-specific details
import handle from "@modelfetch/aws-lambda"; // Choose your runtime
import server from "./server"; // Import your server
export const handler: AWSLambda.LambdaFunctionURLHandler = handle(server); // That's it — ModelFetch handles all runtime-specific details
import handle from "@modelfetch/vercel"; // Choose your runtime
import server from "./server"; // Import your server
const handler = handle(server); // That's it — ModelFetch handles all runtime-specific details
export const GET = handler;
export const POST = handler;
export const DELETE = handler;
import handle from "@modelfetch/cloudflare"; // Choose your runtime
import server from "./server"; // Import your server
export default {
fetch: handle(server), // That's it — ModelFetch handles all runtime-specific details
} satisfies ExportedHandler<Env>;
import handle from "@modelfetch/netlify"; // Choose your runtime
import server from "../server.ts"; // Import your server
export default handle(server); // That's it — ModelFetch handles all runtime-specific details
That's just a few lines of code to make your McpServer
work across all supported platforms.
⚡ The handle()
Function
Every runtime package exports a default handle()
function that takes an McpServer
instance as its first parameter and handles all runtime-specific details:
import handle from "@modelfetch/node"; // Choose your runtime
import server from "./server"; // Import your server
handle(server); // That's it — ModelFetch handles all runtime-specific details
ModelFetch provides runtime-specific packages that handle tedious platform differences while you focus on building your MCP server capabilities. Each package maintains a consistent API across different runtimes.
Package | Description | Status |
---|---|---|
@modelfetch/node |
Run simple MCP servers with Node.js | ✅ Ready |
@modelfetch/next |
Run flexible MCP servers with Next.js | ✅ Ready |
@modelfetch/bun |
Run lightning-fast MCP servers with Bun | ✅ Ready |
@modelfetch/deno |
Run secure MCP servers with Deno | ✅ Ready |
@modelfetch/aws-lambda |
Deploy MCP servers to AWS Lambda | ✅ Ready |
@modelfetch/vercel |
Deploy MCP servers to Vercel | ✅ Ready |
@modelfetch/cloudflare |
Deploy MCP servers to Cloudflare | ✅ Ready |
@modelfetch/netlify |
Deploy MCP servers to Netlify | ✅ Ready |
- Node.js 22+
- pnpm 10+
-
Clone the repository:
git clone https://github.com/phuctm97/modelfetch.git cd modelfetch
-
Install dependencies:
pnpm install
-
Type check, lint, and build all projects
pnpm exec nx run-many -t typecheck lint build
# Run development server for a project
pnpm exec nx dev @modelfetch/website
# Type check all projects
pnpm exec nx run-many -t typecheck
# Lint and auto-fix all projects
pnpm exec nx run-many -t lint --args=--fix
# Build all projects
pnpm exec nx run-many -t build
# Format code
pnpm -w format
- Model Context Protocol - TypeScript SDK
- Model Context Protocol - Documentation
- ModelFetch - Website
- ModelFetch - Documentation
ModelFetch is MIT licensed.
Built with ❤️ by Minh-Phuc Tran