What does agentic commerce mean in simple terms?

It means letting an AI agent do the buying. You tell the agent what you want and set limits such as a budget, and it finds options, decides within those limits, and pays. The term also covers agents buying services for themselves, such as paying per API call.

Is agentic commerce the same as ecommerce?

It is a form of ecommerce where the buyer is an AI agent rather than a person clicking through a site. It needs structured product data, programmatic payments, and ways for merchants to verify that the agent acts with real authority from its principal.

What payment methods does agentic commerce use?

Card-based tokens scoped to a merchant and amount, account-based payment credentials, and stablecoin payments from agent wallets are all used. Stablecoins suit high-frequency, small machine payments, while card tokens suit consumer purchases at merchants that already accept cards. Many systems support both.

Agentic Commerce: Meaning and How It Works

Agentic commerce means commerce in which AI agents act on behalf of a person or business: discovering products or services, comparing options, making decisions within set limits, and completing payment. The human sets goals and constraints; the agent executes. It depends on machine-readable catalogs, delegated payment authority, and protocols that let merchants trust agent-initiated transactions.

The core idea

Traditional ecommerce assumes a human clicks through a storefront and enters payment details. Agentic commerce replaces that human at the point of action with an agent. The agent reads structured product or service data, evaluates options against instructions, and executes the transaction. The human's role moves from operator to principal who defines intent and limits.

Two broad categories

Consumer agentic commerce covers assistants that shop for people: booking travel, reordering supplies, or buying items found in a chat. Machine commerce covers agents buying resources for their own tasks: API calls, data, compute, or other agents' services. The second category often involves many small, frequent payments and no checkout page at all.

How a purchase flows

A typical flow has four steps. The user states intent, such as a budget and requirements. The agent discovers and compares offers through APIs or feeds. The agent presents a choice or proceeds under pre-approved rules. Payment executes through a credential or wallet the agent is authorised to use, and the merchant fulfils the order.

Why payments are the hard part

Card networks and checkout forms were designed for humans with fraud checks tuned to human behaviour. Agents need credentials they can use programmatically without exposing full card details or unbounded wallets. Merchants need proof that the agent acts with real authority. These two needs drive most of the protocol work in the space.

Delegated authority

Delegation means the principal grants the agent a scoped permission: a spending cap, an allowed merchant list, a time window, or a single purchase. Card-based approaches use tokens bound to a merchant and amount. Onchain approaches use session keys or smart account permissions that bound what a key can sign.

Machine-readable supply

Agents cannot use pages designed only for human eyes efficiently. Agentic commerce relies on structured catalogs, product feeds, APIs, and protocols that expose prices, availability, and terms in forms agents can parse and act on reliably.

Trust and verification

Merchants want to know that an agent is legitimate and that its principal authorised the purchase. Approaches include signed mandates, verified agent identities, payment tokens scoped to a transaction, and onchain settlement that is final and auditable.

The protocol landscape

Several open protocols target parts of agentic commerce. The Agentic Commerce Protocol from OpenAI and Stripe focuses on checkout between agents and merchants. Google's Agent Payments Protocol focuses on mandates that prove user intent. x402 uses the HTTP 402 status code for pay-per-request payments, typically in stablecoins.

Where stablecoins fit

Stablecoins such as USDC settle quickly, work at small amounts, and can be held and spent by software wallets without a bank account per agent. That makes them well suited to machine commerce, where an agent may pay fractions of a dollar per API call many times an hour.

Risks to manage

Agents can be manipulated by malicious content, misread instructions, or overspend through loops. Sound designs cap spending, restrict counterparties, log every payment, and give humans a way to revoke authority instantly. Bounded authority matters more than clever prompting.

What businesses should do

Merchants can prepare by exposing structured catalogs and APIs, supporting agent-friendly payment methods, and defining policies for agent-initiated orders. API providers can price per call and accept programmatic payment. Builders of agents need wallets or payment credentials with enforceable limits.

A short history of the term

The phrase gained wide use as large language model agents became capable of multi-step tasks and major payment and technology companies began publishing protocols for agent checkout and payment. It extends earlier ideas such as automated reordering and machine-to-machine payments with agents that can reason about choices.

Agentic commerce versus automation

Rule-based automation, such as a recurring subscription, executes a fixed instruction. Agentic commerce involves an agent choosing among options at runtime, within limits. The distinction matters because decision-making agents need stronger guardrails than fixed schedules.

An example of machine commerce

A research agent needs current pricing data from a provider that charges per request. Instead of a developer creating an account and a subscription in advance, the provider responds to the agent's request with a price. The agent checks the price against its budget, pays a small USDC amount from a bounded wallet, and receives the data, all within one task.

An example of consumer agentic commerce

A user asks an assistant to find a pair of running shoes under a set budget in a particular size. The assistant queries merchants with agent-readable catalogs, compares options, and either asks the user to confirm or buys under a pre-approved rule. Payment uses a credential scoped to that merchant and amount, so the assistant never holds the user's full card details.

Who is building it

Payment networks, payment processors, AI companies, ecommerce platforms, and crypto infrastructure providers are all building parts of agentic commerce. Their work covers agent-readable catalogs, scoped payment credentials, mandate formats that prove user intent, and stablecoin rails for machine payments. Open protocols let these pieces interoperate.

Frequently asked questions

What does agentic commerce mean in simple terms?: It means letting an AI agent do the buying. You tell the agent what you want and set limits such as a budget, and it finds options, decides within those limits, and pays. The term also covers agents buying services for themselves, such as paying per API call.
Is agentic commerce the same as ecommerce?: It is a form of ecommerce where the buyer is an AI agent rather than a person clicking through a site. It needs structured product data, programmatic payments, and ways for merchants to verify that the agent acts with real authority from its principal.
What payment methods does agentic commerce use?: Card-based tokens scoped to a merchant and amount, account-based payment credentials, and stablecoin payments from agent wallets are all used. Stablecoins suit high-frequency, small machine payments, while card tokens suit consumer purchases at merchants that already accept cards. Many systems support both.