We're building our first online shop

We build a lot of things at smartive. What we’ve actually never built: a classic online shop.
Now we are. 🛒
For a new B2B project, we get to build a shop. Product data from the ERP, customer login, customer-specific prices and currencies – the full package.
Only question: how do you even build an online shop in 2026?
Shopify? Salesforce? Build it ourselves? 😬
Okay, building it ourselves was off the table fast. Carts, orders, payments, pricing, and all the other fun commerce problems have been solved often enough.
Big players like Shopify, Salesforce, and the usual suspects were obvious candidates. We also looked at a few headless commerce options – and that’s how we ended up with Medusa.
Medusa is open source, runs on Node.js, and already gives us the commerce basics: products, customers, carts, orders, payments, pricing, inventory, and more.
The more interesting part for us is what Medusa does not prescribe.
The frontend is ours. We can build our own data models, APIs, workflows, and integrations. And if our products come from an ERP, or our B2B shop needs some special snowflake feature, we don’t first have to find the marketplace app that kind of does it.
That fits how we like to build software.
B2B works too
This won’t be a classic B2C shop.
Customers log in, and depending on the customer, different prices and sometimes different conditions apply. Add multiple currencies and an integration with the existing ERP.
Medusa already covers a lot of that. Customer groups, price lists, multiple currencies, sales channels, and other commerce primitives. We can use those as a base and extend them where the project gets specific.
So we don’t have to reinvent commerce. We still get to decide a lot ourselves.
Liking it so far. 🙂
DatoCMS + Medusa + Next.js
For content, luckily, we didn’t have to evaluate all over again.
DatoCMS ❤️
We’ve been using Dato for our websites for a few years now, and we’re still big fans.
So why not combine the two?
DatoCMS does content.
Medusa does commerce.
Next.js does the rest.
Copy, images, landing pages, and SEO content come from Dato. Products, prices, customers, and orders come from Medusa. The Next.js frontend is where it all comes together.
On paper, at least, that feels pretty clean.
And Medusa is clearly into AI too 🤖
Something else stood out during the evaluation: Medusa thinks quite consistently about how software gets built with coding agents today.
And no, we don’t mean yet another AI chatbot in the bottom-right corner.
Medusa ships its own skills for coding agents. There’s also an MCP server that gives Claude Code, Codex, and other agents access to up-to-date Medusa docs and guides.
Then there’s Medusa Cloud: hosting for Medusa projects, plus an agent in the dashboard. It knows your deployments, logs, and the project – and helps when a build or deploy goes sideways.
Medusa calls this AI-native commerce. Big words. 😄
But we like the direction. We’re using coding agents more and more ourselves. If the tools we build with are set up for that: even better.
The vibe has to be right
And then there’s a fairly unscientific evaluation criterion.
Are the people behind it cool?
Sounds a bit dumb. It actually matters to us.
That was already one of the reasons we fell for DatoCMS. Good developer experience, good docs, direct contact, and people who clearly care about their product.
Medusa has given us a pretty similar impression so far.
We had a few questions, wrote to the team, and got fast, useful, technical answers. No endless sales pitch. Just a straightforward exchange with people who know what they’re talking about.
That doesn’t show up in any feature matrix. For a technology you might bet on for the next few years, it can still be decisive.
So: Medusa it is?
That’s what it looks like right now. 🐍
Before this reads like a Medusa ad: we haven’t built a production shop with it yet.
The evaluation was promising. Now comes the interesting part.
Over the next few weeks we’ll start building, and we’ll see how DatoCMS + Medusa + Next.js holds up in a real B2B commerce project.
If it works the way we currently imagine, there’ll be a part 2 with what we learned.
And if we end up on Shopify after all, there’ll probably be a slightly different blog post. 😄

Written by
Raphael Wirth





