A Cloudflare Worker is a JavaScript (or Rust/Wasm) function that runs on Cloudflare’s edge network instead of a server you manage. No VM to patch, no container to build. You write a function, deploy it, and it runs in hundreds of locations. Here’s a real one, deployed in ten minutes.

1. Install Wrangler

Wrangler is Cloudflare’s CLI for Workers.

npm install -g wrangler
wrangler login

wrangler login opens a browser tab to authorize the CLI against your Cloudflare account.

2. Scaffold a project

npm create cloudflare@latest my-worker -- --type hello-world
cd my-worker

This gives you a minimal src/index.js (or .ts) and a wrangler.toml config file.

3. Write the endpoint

Open src/index.js and replace it with something real — an endpoint that returns JSON based on the request:

export default {
  async fetch(request) {
    const url = new URL(request.url);

    if (url.pathname === "/api/hello") {
      return Response.json({
        message: "hello from the edge",
        colo: request.cf?.colo ?? "unknown",
      });
    }

    return new Response("Not found", { status: 404 });
  },
};

request.cf.colo tells you which Cloudflare data center handled the request — useful for confirming it’s actually running at the edge, not just locally.

4. Run it locally

wrangler dev

This starts a local server that emulates the Workers runtime (via workerd, the same engine Cloudflare runs in production — not a Node polyfill). Hit http://localhost:8787/api/hello and confirm you get JSON back.

5. Deploy

wrangler deploy

You’ll get a URL like https://my-worker.<your-subdomain>.workers.dev. That’s live, globally, immediately — no build pipeline, no waiting for a CDN to propagate.

What you don’t get for free

• No filesystem. fs doesn’t exist. State goes in KV, D1, R2, or Durable Objects.
• CPU time limits. Free tier caps CPU time per request (currently well under a second of actual compute); this is fine for API glue and header logic, not for image processing or heavy loops.
• No persistent in-memory state between requests unless you use a Durable Object — each request may hit a different isolate.

The failure you’ll hit first

Importing a Node-only package:

Error: No such module "node:crypto".

Workers ship a node_compat flag for partial compatibility, but not every package works. Check before you build a feature around a library — search for “workers compatible” or check if it depends on native bindings.

Next step

Bind a KV namespace or D1 database once you need state beyond a single request — plain fetch-and-respond Workers like this one cover routing, auth checks, and header rewrites without touching storage at all.