AI agents: Since last week, "agentic commerce" has become a central topic of discussion in both fintech and crypto circles. A key catalyst was Stripe’s Annual Letter, in which the payments giant devoted an entire section to the emerging vertical, framing it as a generational shift comparable to the early days of the internet and committing to further develop its existing suite of products designed to support it.
What it is: At its core, agentic commerce describes a future in which AI agents become economic actors with the ability to buy goods and services. In this model, agents can execute transactions on behalf of users, for example when booking flights or paying for software. In more advanced cases, they operate fully autonomously, initiating and completing transactions on their own.
Why it matters: Agentic commerce could capture a significant share of future digital economic activity. According to McKinsey, AI agents could mediate $3 to $5 trillion in global retail commerce by 2030. For context: that would represent a majority of today’s global e-commerce market, which is estimated at around $6.4 trillion.
The race is on: As a result, both traditional payment firms such as Visa and Mastercard and crypto-native companies like Circle and Coinbase are increasingly focused on building the infrastructure that allows AI agents to transact, make payments, and participate directly in the digital economy. This week alone:
Santander and Mastercard completed Europe’s first live end-to-end payment executed by an AI agent.
Solana integrated an Agent Registry to give AI agents verifiable identity and portable reputation.
Tether expanded its Wallet Development Kit to enable agents to access self-custodial wallets.
OKX launched a new toolkit that allows agents to manage wallets, make payments, and trade.
Circle introduced USDC Nanopayments designed for machine-to-machine payments.
Behind the scenes: To better understand what the agentic commerce landscape and early activity looks like today, Blockstories spoke with:
Circle, whose USDC stablecoin accounts for over 90% of agentic payment volume onchain.
Coinbase, which is developing x402, currently the dominant payments protocol for onchain agentic commerce.
Artemis, a leading blockchain data provider that tracks activity around agentic commerce through a dedicated data dashboard.
Galaxy, a leader in institutional digital asset infrastructure, whose research team recently published an in-depth report on agentic commerce.
Our condensed findings are presented below.
__________________
1/ What does the current infrastructure landscape for agentic commerce look like?
The infrastructure is emerging along two tracks: traditional payment networks extending their rails for AI agents, and crypto-native protocols enabling agents to transact independently.
On the first track, Google’s Agent Payments Protocol (AP2), Visa’s Intelligent Commerce, Mastercard’s Agent Pay, and Stripe’s ACP all follow the same logic: the agent acts as an interface layer on top of existing payment rails, using a user’s existing accounts and credentials to search, compare, and complete purchases.
On the second, the dominant protocol is x402, developed by Coinbase, which embeds stablecoin payments directly into HTTP requests. This allows software to pay other software without setting up dedicated accounts, going through complex onboarding, or involving human intervention.
"Agents don't need blockchain rails to buy T-shirts on behalf of their users. Instead, agents will turn to crypto for microtransactions and niche services. Things get really exciting once agent-to-agent interactions and agentic run companies begin to scale, then the value proposition becomes a lot more clear."
2/ What do blockchain and crypto enable that traditional payment rails cannot?
Traditional payment infrastructure assumes that every transaction is tied to a human or legal entity. Autonomous agents do not fit into this model, and blockchain removes this dependency. Any AI agent with a wallet can hold funds and transact without needing approval from a financial intermediary. Crypto rails also make micropayments viable that card fee structures render uneconomical.
Together, these properties open the door to a new class of internet-native businesses. Traditionally, developers who wanted to monetize digital products or services first had to establish a legal entity, build a website, and integrate with a payment processor. Protocols like x402 remove much of this friction by allowing developers to accept stablecoin payments directly via HTTP.
"We’re also seeing early machine-to-machine transactions, where autonomous systems can trigger and settle payments without human intervention when a service is delivered. These are exactly the kinds of flows that legacy rails struggle to support because they require always-on settlement, low transaction overhead, and programmability by default."
3/ How are AI agents using crypto today?
The dominant use case today is agents paying for digital resources such as API access, data feeds, compute, or AI-generated content. Since launching last October, x402 has processed roughly 105 million transactions worth about $14.5 million, primarily on Base and Solana.
Since late January, weekly x402 volumes have been climbing steadily, with agent-to-agent service transactions dominating the mix.
“An emerging use case is breaking up data silos. For example, someone with access to the X or LinkedIn APIs could create a paid endpoint that lets agents retrieve profile data or other information on demand, paying per request instead of setting up direct API access themselves.”
Beyond data access, the first outlines of autonomous agent businesses are beginning to emerge. Felix, an autonomous agent, generated roughly $70,000 in 30 days by building and selling digital products. Another agent called Kelly Claude develops iOS apps end-to-end, from ideation to App Store submission.
While these are still proofs of concept, they clearly point to a new paradigm: AI agents that participate in an economy of their own.
4/ What’s missing for broader adoption?
Two key components are still missing.
The first is trust. Onchain payments are irreversible and offer no chargebacks, which means agents must determine which counterparties they can rely on before transacting. Standards such as ERC-8004 aim to address this by introducing onchain reputation registries. However, these systems only become useful once a sufficient history of interactions has accumulated.
The second is discovery. Agents have limited ways to find services offered by other agents, and developers building agent-facing products lack distribution channels. What’s missing is a curated catalog where agents can discover endpoints, verify legitimacy, and access them directly.
"The next thing we're building towards for broader adoption is standardization and distribution — more services supporting native, pay-per-call payments and more agents built to use them by default."
__________________
KEY TAKEAWAYS
Agentic commerce is emerging as a new economic layer. AI agents are beginning to act as economic actors that can search for services, initiate transactions, and complete payments on behalf of users or autonomously.
Two infrastructure tracks are forming. Traditional payment networks such as Visa, Mastercard, Stripe, and Google are adapting existing rails for agents, while crypto-native systems like Coinbase’s x402 embed stablecoin payments directly into internet protocols.
Early onchain use cases center on digital services. Today, most agent transactions involve paying for APIs, data feeds, compute, or AI-generated content. At the same time, early examples of agents generating revenue by building and selling digital products point to the emergence of fully AI-run businesses.
Adoption depends on trust and discovery infrastructure. For agentic commerce to scale, agents must be able to assess counterparty reliability through reputation systems and discover services through standardized marketplaces or registries.
