Anthropic brings Claude into commerce with Visa and Mastercard

Anthropic has launched a blueprint enabling retailers to build AI-powered shopping and merchant agents using Claude, with support from Visa, Mastercard and Accenture.

Anthropic brings Claude into commerce 

The framework is aimed at retailers, travel companies, telecoms providers and ticketing businesses, allowing them to embed Claude-powered agents directly into their own digital channels rather than surrendering the customer relationship to a third-party shopping platform.

For the payments industry, however, the most important point is what the agents do not yet do: Anthropic’s shopping agent can discover products, compare options and construct a basket, but the final transaction remains with the merchant’s existing checkout or an external payments provider.

AI moves up to the point of payment

The customer-facing agent can search a merchant’s catalogue using natural language, remember preferences, answer product questions, make recommendations and assemble an order.

That brings AI much closer to the payment event, but stops short of allowing Claude itself to initiate and complete a purchase autonomously.

This distinction matters because agentic commerce introduces a new set of payments requirements. If an AI system is eventually permitted to transact for a consumer, issuers and payment networks will need to establish that the agent is authorised, that the transaction falls within the customer’s mandate and that credentials can be presented securely without exposing sensitive card data.

That is where Visa and Mastercard become strategically important.

Both networks are developing agentic-commerce frameworks intended to support authenticated AI transactions, including controls around identity, payment credentials, merchant acceptance and consumer consent. Their role is therefore less about powering Anthropic’s conversational interface and more about creating trusted payments infrastructure beneath it.

Merchant agents stay outside transaction execution

Anthropic has also introduced a merchant-facing agent capable of monitoring inventory, analysing sales, recommending pricing changes and proposing promotions.

These recommendations require human approval before implementation, an important safeguard given the commercial and regulatory consequences of allowing AI systems to alter prices or customer offers autonomously.

Shopify and Priceline are among the businesses already using Claude-powered commerce technology, while Anthropic says retailers deploying its agents have recorded larger baskets and higher conversion rates.

Agentic commerce creates a new payments control layer

The larger payments question is what happens when AI progresses from influencing a purchase to authorising one.

At that point, the industry will need mechanisms for agent identification, tokenised credentials, transaction limits, authentication, revocation, liability and dispute resolution. Merchants will also need confidence that an AI-generated order represents genuine consumer intent rather than an erroneous or manipulated instruction.

That makes agentic commerce less a replacement for established payments infrastructure than a new layer sitting above it.

Anthropic is building the interface through which consumers and merchants may increasingly make purchasing decisions. Visa, Mastercard, issuers and payment service providers are positioning themselves to determine how those decisions are converted into authorised, secure and accountable transactions.

For payments companies, that is the real significance of the launch: the interface may increasingly belong to AI, but the economics, risk and trust framework of the transaction will still be decided in the payments layer.