Concept
What is agentic commerce?
The term gets used loosely, so this page pins it down: the two things people mean by it, the practical version that exists now, and what is still emerging.
Short answer
Agentic commerce is commerce where software agents can participate directly in buying or operating a store. The practical version available today is seller-side: a store whose core actions, such as creating products, publishing checkout, reading orders, and issuing refunds, are callable by software instead of being trapped inside a human dashboard.
The payments, products, and orders underneath are still ordinary commerce. What changes is the actor. When an agent is the one creating a product, publishing checkout, or issuing a refund, the commerce system has to expose those actions in a form software can call safely.
Two senses of the term
People use agentic commerce to mean two different things. They are related but not the same, and conversations get muddled when nobody says which one they mean.
Selling through AI agents
Here the agent is the buyer. A person asks an assistant to find and purchase something, and the agent completes the transaction. This is the buy side. It depends on agents being able to discover products and pay, and on stores being reachable in a way an agent can act on. This sense is the one still settling, since the standards for agent-driven purchasing are early.
A store that is agent-operable
Here the agent runs the store. Instead of a human in a dashboard, an agent creates products, publishes checkout, reads orders, and issues refunds through an API or a tool interface. This is the sell side. It is more concrete, because it only requires the store's own actions to be callable by software, and that is buildable today.
The buy side needs the whole internet to agree on how agents shop. The sell side needs one store to expose its operations cleanly. That is why the seller-side definition is the useful one today.
Where commerce.fyi sits
commerce.fyi works on the sell side. It turns a description into a finished store and exposes that store's operations so a person or an agent can run it. The same commerce actions are reachable two ways:
/dev
Start here for API, MCP, llm.txt, the skill surface, and docs.
REST
HTTP routes for stores, products, checkout, orders, refunds, and receipts.
MCP
Tools that let agents create, operate, and inspect stores.
SDK / CLI
Planned, not available today.
The store actions themselves are concrete. An agent or a developer can call each one today:
Create a storefront from a short description.
POST /v1/storesAdd a product. Description and image are optional.
POST /v1/productsUpdate typed storefront copy: name, tagline, and hero copy.
PATCH /v1/stores/{id}Update a product name, price, description, or metadata.
PATCH /v1/products/{id}Create a hosted Stripe checkout URL for the buyer.
POST /v1/checkoutsList paid, pending, and refunded orders for a store.
GET /v1/orders?store_id=...Refund a paid order in full or partially.
POST /v1/orders/{id}/refundSend a receipt email through Resend. Idempotent.
POST /v1/notifications/receiptsThat is the agent-operable store in practice: create a store, add products, publish checkout, read orders, refund, and send receipts. Real Stripe checkout sits behind it, and every order, refund, and receipt is a real record.
Shipped versus emerging
Agentic commerce attracts big claims, so it is worth being plain about what holds up.
Shipped today
Creating a store from a description, adding and updating products, publishing checkout, reading orders, issuing refunds, and sending receipts, all reachable over REST and MCP. A person or an agent can do this work now against a store with real Stripe checkout.
Still emerging
Fully autonomous agents that shop across many stores without a human in the loop, and the shared standards that would let them, are early and unsettled. An agent running a store still publishes and operates under a person's direction. Treat claims of hands-off autonomous commerce with caution; the plumbing for it is not finished.
The honest read: the sell side of agentic commerce is real and usable, the buy side is promising and unfinished. A store that an agent can operate is something you can build this week. A web where agents shop freely is something the industry is still working out.
Why this matters for sellers
If agents are going to participate in commerce, the stores they reach have to expose real operations, not just web pages. A store whose every action is callable is ready for that shift, and it is also easier for the seller to automate today. The agent-operable store is useful now and positioned for what comes next.
That is the bet commerce.fyi makes: build the store so it can be run by software, keep the commerce records real, and let a person stay in control of what publishes.
You know what you sell. Get the store.
No signup to start. Describe your products and see a real store you can take live.