Let your customer's AI buy directly.
Agentic commerce isn't the future. We don't wait for the platforms and the AIs to open this up — we build it for your store now. Two things on top of your data: a shopping assistant that closes the sale in chat, and your store's MCP endpoint through which a customer's own AI buys.
We do this while others wait
A native agentic checkout exists on two platforms (Shopify, WooCommerce). Whether others follow, and when, is uncertain.
The EU does not allow an AI agent to buy entirely on its own: PSD2 and SCA require human approval.
AIs cannot yet auto-discover a store's MCP from its data. That standard is maturing, we estimate late 2026.
Until then the same is achieved with a connector. Either way: ready today.
The customer asks, the AI sells
The assistant asks what's needed (“vegan, nut-free gift under €30”), recommends the right product from the enriched data and closes the sale, all in one conversation.
Payment runs through Stripe inside the conversation. The pilot produced a real order in the store.
Claude connects to the store's MCP over the public internet. Order #69 in the pilot.
The future picture: the AI finds the store's endpoint from its own data. We are already ready for it.
How your customers start buying
You bring customers to your store's AI checkout endpoint, we provide the tools. It fits repeat buyers, subscribers and B2B customers especially well.
Send a “buy from us with your AI” link. “Order your next one with your AI agent” — one glance and it's added.
A card in the box and a discount code for the first AI order.
“Order my coffee again” — the customer's own AI handles it.
A professional buyer specifies and orders with the AI they use at work.
Why a customer would add your store
“Order my usual coffee” — done. No re-picking, no filling in details.
The AI knows your products, sizes, prices and availability. No wrong store, no hallucinated products.
Like your best salesperson inside the customer's AI — with an offer only for them if you want.
The purchase happens where the customer already is. No new app, no new tab.
In B2B this matters even more
“Order 200 units of SKU-4471 to the Vantaa warehouse.” Right in the Claude the buyer already works in, with contract pricing and account known.
Connect the supplier's MCP and your own stock levels: the buyer's AI prepares a replenishment order when stock drops below the limit. A human approves.
An engineer specs a project with AI, the supplier's deep product data surfaces the right part with its compatibilities, and the order happens in the same conversation.
We build it on an open commerce API
The agent needs three things from a platform: reading products, creating an order, and updating payment status.
- Piloted
- WooCommerce — a working end-to-end implementation.
- Buildable
- All mainstream platforms, including the Finnish MyCashflow, Vilkas and Finqu. The exception is closed website-builder services with no commerce API.
- Enterprise
- SAP Commerce Cloud, Salesforce Commerce Cloud, commercetools: project-specific, but doable.
- Payments
- Native agentic checkouts are card-first. In Europe that leaves money on the table, so we handle local payment methods.
Without product data, none of this works
An agent is only as good as the store's structured, contextual product data. That's why the agent layer always sits on top of the data layer — and data depth is exactly what almost every store is missing.
Could an AI buy from you?
Don't take our word for it. Paste this into your own Claude or ChatGPT and ask it to attempt a purchase from your store. It won't be able to finish, but it will tell you exactly where it stops.
Act as a buying AI agent. Your customer wants: . The store is . Try to take the purchase as far as you can using public information only. Do not guess anything — if information is missing, say it is missing. Then report three things: 1. Where you stopped and why. 2. Which product details you would have needed to choose the product with confidence. 3. What the store would have needed to expose for you to place an order without a human. Do not submit any forms and do not attempt any payment.
The list you get back is exactly the work the agent layer does. We build it for you, on top of the data layer. Book a demo