Medusa

Last reviewed Aug 26, 2026 · 770 words

Installing Medusa gives you a working admin dashboard, a REST API, and no shop. There is nothing a customer can visit until you deploy a storefront, and the official Next.js starter is a starting point rather than a finished theme. That is the whole decision in one sentence: Medusa is for a team that wants to write a storefront in TypeScript, and WooCommerce is for one that wants to pick a theme and start selling this week.

What the install actually produces

npx create-medusa-app@latest scaffolds a backend project, asks for a Postgres connection, runs migrations, and seeds an admin user. The backend listens on port 9000 and serves the admin at http://localhost:9000/app. From there you can create products, regions with their own currencies and tax settings, shipping options, and discount codes. The API is documented and complete; a headless frontend on any framework can consume it.

Optionally the installer also clones the Next.js starter storefront, which runs on port 8000 and talks to the backend with a publishable API key. It has product listings, a cart, checkout, and account pages, styled plainly. Treat it the way you would any boilerplate: everything a real store needs beyond that (your brand, search, reviews, content pages, analytics) is your work.

Postgres is required, Redis is expected

Postgres is the only supported database. Redis is optional in development and effectively required in production, where it backs the event bus, caching, and workflow engine; without it Medusa falls back to in-memory modules that lose state on restart. Budget the 1 GB of RAM in the listing for the Node process alone and add Postgres and Redis on top, so a 2 GB VPS is the practical minimum and 4 GB is where I would start for anything with real traffic. A Docker deployment means building the backend image yourself from the project plus standard postgres and redis containers; there is no official all-in-one image, which surprises people arriving from WordPress.

Operationally the backend is an ordinary Node process. npm run build produces the bundle, npm run start serves it, and the environment carries DATABASE_URL, REDIS_URL, JWT_SECRET, COOKIE_SECRET, and the STORE_CORS, ADMIN_CORS, and AUTH_CORS origins for the domains your storefront and admin live on. Get the CORS values wrong and the admin loads but every request fails, which is the first thing to check. Backups are a Postgres dump plus whatever file storage you configured for product images; the local file provider writes to the project's static directory by default, so mount it.

The module system is the point

Medusa v2 splits commerce into modules (product, pricing, inventory, cart, order, fulfilment, payment, tax) that talk through a shared workflow engine. You replace or extend them individually: use the maintained Stripe provider for payments, add a custom fulfilment provider for your courier, or write a module that syncs stock to a warehouse system. Multi-region pricing and multi-currency are built into the pricing and region modules rather than bolted on, which is where Shopify's cheaper plans make you pay extra, and it is the strongest point in the Shopify alternatives discussion for a business selling into several countries.

The honest cost is that every one of those modules is code. There is a plugin ecosystem, but it is thinner than WordPress's and parts of it lag a version behind, because the v1 to v2 migration reset it.

Who wins with it, and who does not

A development agency building a bespoke store, a company with an existing frontend that needs a commerce backend, or a marketplace with rules no template supports: Medusa is a strong fit and its MIT license means no revenue share ever. A small shop owner without a developer will be miserable inside a week. Between those, Saleor is the closest comparison in the ecommerce category: GraphQL-first, Python, a more opinionated dashboard, and the same bring-your-own-storefront reality.

What I'd do

If you have a TypeScript developer on the team and a storefront you want to own completely, run Medusa on a 4 GB VPS with Postgres and Redis alongside, deploy the Next.js starter behind a reverse proxy, and plan 2 to 4 weeks of frontend work before launch. If nobody on the team writes code, WooCommerce on managed WordPress hosting gets you selling faster, and the money saved on development pays for a lot of plugins. Do not pick Medusa because it is newer; pick it because you were going to write the frontend anyway.