What is agentic commerce?

Software is starting to do more than recommend. It buys, sells, and settles on its own. Agentic commerce is the name for that shift, and the payment rails it runs on look very different from a card form.

← All posts

A working definition

Agentic commerce is commerce carried out by autonomous software agents rather than by a person clicking buttons. An agent is given a goal, and to reach it, it takes actions in the world: it calls APIs, buys data, subscribes to services, and pays for compute, all without a human approving each step. When those actions involve money, the agent needs a way to pay that does not assume a human is present.

That last part is the whole story. For thirty years, online payments have been designed around a person: a checkout page, a card number, a one-time password on a phone. An agent has none of those. It needs to pay programmatically, in small amounts, at machine speed, within limits its owner set. That requirement is reshaping payments from the ground up.

Why now

Three things arrived at once. Models got good enough to plan and act over many steps. Tool-use and standard protocols gave agents a reliable way to call external services. And stablecoins matured into fast, dollar-pegged money that moves in seconds for cents. Put together, an agent can now reason about what it needs, find the service that provides it, and pay for it, in one continuous loop.

The three things an agent needs to transact

  • A wallet of its own. Not your personal account, but a scoped wallet with a balance and hard spending limits, so a mistake or a bad instruction cannot drain everything.
  • A way to be charged per action. Traditional billing assumes contracts and invoices. Agents need to pay for a single API call the moment they make it. The x402 standard, built on HTTP 402, does exactly this.
  • Money that settles instantly and globally. Stablecoins like USDT and USDC settle in seconds on-chain, with no borders and no banking hours, which is the only thing that keeps pace with an agent.

What it looks like in practice

Imagine a research agent asked to compile a market report. It calls a paid data API and hits a 402 asking for two cents; it pays from its wallet and continues. It needs a premium search; it pays per query. It runs a large inference on a paid model; it settles that too. At the end, its owner has a report and a complete on-chain record of every cent it spent and where. No card forms, no monthly invoices, no surprises, because the wallet enforced the budget the whole time.

The shift is from "a human buys a subscription so software can act" to "software pays for exactly what it uses, as it uses it."

What it means for your business

There are two sides to prepare for. If you build agents, they will need spend-capped wallets and a way to pay for services safely. If you sell a service, an entirely new class of customer is arriving: agents that can pay you directly, per call, with no sales cycle. Making your API payable via x402 means an agent can discover it, use it, and pay for it without a human ever filling out a form.

Either way, the rails are stablecoins plus per-call payments plus scoped wallets. That is what Sendbase is built for. See AI agent wallets for the spend-cap and allowlist model, and x402 payments for how per-call charging works.

The bottom line

Agentic commerce is not a far-off idea; the pieces are shipping now. The businesses that get ahead of it will be the ones that give their agents safe wallets and make their services payable by machines. Start by giving one agent a spend-capped wallet and letting it pay for one real thing. Everything else builds from there.

Build for agentic commerce

Open a free, self-custodial account, give an agent a spend-capped wallet, and let it pay per call.