top of page

Coinbase's AiFi Just Put a Price on the AI Economy: What Agentic Payments Mean for Fintechs

Aug 18
8 min read

Coinbase’s move into AiFi : agentic finance : is more than another crypto product launch. It is a signal that the next major contest in payments may not be fought at the checkout screen, inside the banking app or even across the card network.

It may be fought inside the software layer where AI agents decide what to buy, which merchant to use, how much to pay and which payment rail should settle the transaction.

Coinbase has launched an AiFi financial stack built around Base, its Ethereum layer-2 network, USDC and the x402 open payment standard. The company says agent transactions using x402 have surpassed US$100 million in volume.

That milestone matters because it demonstrates that agentic payments are moving from a theoretical use case to an emerging transaction category. It also shows Coinbase attempting to move beyond its traditional role as a crypto exchange and become part of the infrastructure through which autonomous software holds, moves and spends money.

For fintechs, payment companies, Payfacs, issuers, POS providers and remittance businesses, the strategic question is no longer whether AI agents will participate in commerce.

The question is: who will control the agency layer between human intent and the movement of money?

What is Coinbase’s AiFi stack?

Coinbase’s AiFi strategy brings together several components:

  • Base for low-cost, on-chain transaction execution

  • USDC as a programmable settlement asset

  • x402 for machine-to-machine payments

  • Agent wallets that allow software agents to hold and use funds

  • Developer tools that connect agents to payments, APIs and on-chain applications

  • Business acceptance tools for merchants and platforms

The x402 standard uses the existing HTTP 402 “Payment Required” status code to create a machine-readable payment flow.

In a simplified example, an AI agent requests access to an API, dataset or digital service. The service responds with payment requirements. The agent assesses those terms, authorises a payment from its wallet and retries the request with proof of payment. The service is then delivered.

This is fundamentally different from a conventional card checkout. There is no requirement for an agent to create a consumer account, enter card details, navigate a multi-step form or wait for a human to approve every small transaction.

The result is a payment model designed for software-to-software commerce, including pay-per-use APIs, data access, compute, digital services and eventually physical goods.

Coinbase has positioned Base, USDC and x402 as the core of this infrastructure. According to the company’s public communications and related reporting, x402 has already processed more than US$100 million in transactions, alongside significant machine-to-machine activity on Base.

The important point is not only the size of the number. It is the type of transaction it represents: frequent, automated and initiated by non-human actors.

Why the US$100 million milestone is strategically important

US$100 million is modest compared with global card or account-to-account payment volumes. But comparing it directly with those markets misses the point.

Agentic payments are still establishing their infrastructure, standards and trust models. At this stage, volume is evidence of product-market formation, not market maturity.

The milestone suggests three things.

1. Agents are becoming economic participants

AI agents are no longer limited to generating text, making recommendations or searching for information. They are beginning to interact with services that require payment.

That creates a new customer category: the machine buyer.

A machine buyer may purchase an API call, a data licence, cloud compute, a transport booking, a digital subscription or inventory on behalf of a person or organisation. Its payment behaviour will be different from a human customer’s behaviour.

It may transact more frequently, use smaller values, compare multiple rails in real time and require continuous rather than one-off authorisation.

2. Stablecoins have a credible role in machine payments

USDC is useful in this model because it can move programmatically, operate across borders and support small-value transactions without relying entirely on legacy billing systems.

That does not mean stablecoins will replace cards or bank payments. It does mean they are becoming a serious option for specific machine-native use cases, particularly digital services, cross-border settlement and agent-to-agent transactions.

For remittance businesses and international payment providers, this is particularly relevant. An agent may eventually compare settlement speed, foreign exchange cost, liquidity and regulatory constraints before selecting a payment method.

3. The infrastructure market is opening early

Coinbase is not simply offering agents a wallet. It is attempting to provide the surrounding stack: money, execution, payment messaging, developer access and merchant acceptance.

That is a deliberate infrastructure strategy.

The company is effectively saying that the future opportunity is not only helping people buy crypto. It is helping autonomous software hold value, make decisions and execute transactions.

Abstract digital illustration showing a layered AI agentic payments stack from identity and permissions to settlement

The agency layer is the real battleground

The industry often describes agentic commerce as a new form of checkout. That framing is too narrow.

In traditional commerce, the customer usually controls the journey. They search, compare, select a merchant, choose a payment method and authorise the transaction.

In agentic commerce, the agent may control most of those steps. The human defines the objective and the boundaries, while the agent executes the process.

That creates a new layer of commercial influence.

The organisation that controls the agency layer may influence:

  • Which merchants are discovered

  • Which offers are presented

  • Which payment methods are preferred

  • How credentials are stored and tokenised

  • Which risk rules are applied

  • How disputes and refunds are managed

  • Which data is shared with merchants and payment providers

This is why Coinbase’s AiFi strategy should be viewed alongside initiatives from the broader payments ecosystem.

  • Stripe’s Agentic Commerce Protocol (ACP) focuses on agent-to-merchant commerce and checkout flows.

  • Stripe’s Machine Payments Protocol (MPP) provides an open framework for machine-initiated payments, including microtransactions and recurring payments.

  • OpenAI ACP is closely associated with the agentic commerce protocol developed with Stripe and designed for agent-led purchasing experiences.

  • Visa Agentic Ready is helping issuers test and validate agent-initiated transactions within controlled production environments.

  • Mastercard Agent Pay aims to extend Mastercard’s network capabilities into agentic commerce.

  • Google UCP addresses product discovery, checkout, identity linking and order management across agentic journeys.

These initiatives are not identical. Some focus on discovery and checkout. Others address authorisation, trust, credentials or settlement.

Together, however, they reveal the competitive structure of the market: technology companies, payment networks, crypto platforms and financial infrastructure providers are all trying to become essential to the agent’s decision-making process.

What fintechs should take from the protocol race

The emergence of multiple protocols does not mean every fintech should immediately build support for every standard.

It does mean payment leaders need a deliberate protocol strategy.

Fintechs and payment service providers

The first priority is making core payment capabilities available through APIs that agents can understand and use.

That includes pricing, payment method availability, settlement timing, refund rules, transaction status and risk requirements. If an agent cannot evaluate a provider quickly and reliably, that provider may not be selected.

Payment providers should also assess whether they can support:

  • Agent identity and authentication

  • Delegated permissions and spending limits

  • Tokenised credentials

  • Real-time risk decisioning

  • Machine-readable receipts

  • Automated reconciliation

  • Agent-specific dispute handling

Payfacs, ISOs and POS companies

For Payfacs and POS providers, agentic commerce will not remain purely digital.

Agents may place orders, manage replenishment, negotiate supplier purchases or select services for businesses. The POS and merchant platform will need to expose product, inventory, pricing and fulfilment data in ways that agents can access.

This makes the merchant operating system increasingly important. The provider that owns the merchant’s data and workflow may become the bridge between the agent and the payment network.

Issuers and remittance businesses

Issuers need to decide how their credentials will be used when the cardholder is not directly present at the point of decision.

That requires more than approving a tokenised transaction. Issuers will need clear controls for:

  • Who delegated authority to the agent

  • What the agent is allowed to purchase

  • Which merchants or categories are permitted

  • How unusual behaviour is detected

  • How the customer can pause or revoke access

Remittance businesses should also consider how agents will compare fiat, stablecoin and local payment options. An agent may optimise for total delivered value rather than simply selecting the customer’s default payment method.

Trust will determine the speed of adoption

The most difficult problem in agentic payments is not moving money. It is proving that the agent was authorised to move it.

Accenture estimates that more than 30% of online commerce could run through AI agents by 2030, representing close to US$3.1 trillion in transactions. But the same market cannot scale on weak controls.

An agentic payment needs a clear chain of intent:

  1. A person or business defines the objective.

  2. The authority is translated into permissions and constraints.

  3. The agent acts within those limits.

  4. The payment provider verifies the agent, credentials and transaction context.

  5. The transaction is recorded in a way that supports audit, reconciliation and dispute resolution.

Fraud will also evolve. Instead of only impersonating customers, criminals may impersonate agents, manipulate prompts, compromise wallets or exploit automated merchant workflows.

The control layer will therefore become a major source of competitive differentiation. Payment providers that can offer identity, consent, risk management and auditability may become more valuable than those competing on processing alone.

Futuristic illustration of multiple payment gateways competing around a central AI decision engine

The strategic choice for payments leaders

RivaTech’s view is that payment businesses should consider three possible positions in the agent economy:

Own the agent

This is the most visible position, but also the most difficult. It requires customer relationships, relevant data, reliable orchestration and a reason for users to delegate decisions to your agent.

Become the agent’s preferred payment option

This means competing on the factors an agent can evaluate: cost, speed, acceptance, reliability, rewards, risk outcomes and settlement certainty.

Own the control and infrastructure layer

This may be the most defensible position for many fintechs. It includes delegated authorisation, identity, tokenisation, multi-rail routing, fraud controls, compliance, reconciliation and dispute management.

Coinbase’s AiFi launch makes the strategic direction clear. The future of payments will not be defined only by how humans pay. It will also be defined by how software agents select, authorise and execute payments.

For fintechs, the opportunity is to prepare before the agent becomes the default customer interface.

That means reviewing your API strategy, merchant data, wallet architecture, risk model and payment-rail economics now. RivaTech’s payment industry consulting services can help payment businesses assess these capabilities across cards, alternative payments, cryptocurrency, softPOS and digital payments.

The companies that win will not simply accept transactions from agents. They will become trusted, visible and selectable inside the workflows where agents make economic decisions.

For more analysis, read RivaTech’s coverage of agentic commerce and the US$3 trillion opportunity and the emerging machine wallet.

Frequently asked questions

What is Coinbase AiFi?

Coinbase AiFi, or agentic finance, is a financial infrastructure strategy for AI agents. It combines Base, USDC, x402, agent wallets and developer and business tools so autonomous software can hold funds and make payments.

What is x402?

x402 is an open payment standard that uses the HTTP 402 “Payment Required” response to support machine-to-machine payments. An agent can receive payment instructions, authorise a transaction and provide proof of payment before accessing a service.

Why are agentic payments important for fintechs?

AI agents may increasingly control product discovery, payment selection and checkout. Fintechs that are not visible and accessible within agent workflows risk losing influence over customer acquisition, payment choice and transaction data.

Will agentic payments replace cards?

Not necessarily. Cards, bank payments and stablecoins are likely to coexist. Agents may choose between them based on cost, speed, rewards, acceptance, risk and the user’s preferences.

What should payment businesses do now?

Payment businesses should develop an agentic commerce strategy covering API access, delegated authorisation, identity, tokenisation, risk, multi-rail payments, reconciliation and dispute management.

 
 
bottom of page