Last Friday, I was at a barbecue with an OpenClaw maintainer discussing which payment track agents should use. The debate lasted only about 30 seconds before he declared, "Anything but cryptocurrency will do." This wasn't entirely without merit. Following the Moltbook incident and a wave of token launches flooding forums, the OpenClaw Discord community had completely banned mentions of cryptocurrency. Every serious discussion about agent architecture inevitably ended up being linked to some token code. Therefore, banning the topic was a reasonable response. This negative association led serious non-crypto developers to view "cryptocurrency" as a burden, not infrastructure. If you're launching an AI product in 2026, the last thing you want is for your payment layer to be burdened with speculation. However, this reflexive aversion masks an important fact: stablecoins are not speculative tokens. They don't have "launch events" or go to zero. They are a measure of value, settled without banks, transferred without wire transfers, and cost only a fraction of a cent per transaction. Peter Steinberger's decision to integrate OpenClaw with CreditClaw (a virtual card solution powered by Stripe) immediately drew criticism from the crypto community. In 2026, many considered relying on traditional bank card infrastructure a step backward, accusing him of misdirection. However, for his use case, he was right. Steinberger's choice of bank cards was correct because of their obvious advantages. The bank card payment track is indeed the right choice when an intelligent agent acts as a human agent to perform tasks that humans want to do but are unwilling to do themselves. The design intent of bank card networks is very clear: to enable people to spend their own money. Users open accounts with issuing banks, which provide bank cards and grant credit limits or deduct balances. The bank card network routes transactions between issuing banks and merchant acquiring banks. The network and issuing banks are responsible for enforcing policies, rules, and final settlement. Most notably, they employ network tokenization technology, replacing the original card number with a merchant-specific token. These tokens serve as a secure alternative to the underlying cards, preventing them from being used elsewhere (which is extremely helpful for agent control). These networks are deeply optimized for buyer protection. Chargebacks, fraud revocation, and dispute resolution mechanisms have been refined over decades and are now operational on a large scale globally.If an agent books a flight and the airline goes bankrupt, the buyer has recourse. If a stolen card is used for purchases, the issuing bank can reverse the fraudulent transaction. If a consumer buys a laptop online but never receives it, they can dispute the charge and get their money back. This layer of security is the foundation of the bank card system. Another advantage is merchant acceptance. Visa and Mastercard are accepted at over 150 million locations worldwide. In contrast, stablecoins have negligible acceptance, primarily limited to crypto-native merchants and specialized payment processors. For any agent performing tasks within the existing economic system—such as booking travel, ordering office supplies, paying SaaS bills, or purchasing software licenses—bank cards are clearly the winner. Finally, there are consumer incentives: bank card networks possess a psychological and economic lever that stablecoins currently lack—consumer rewards. In the US and other mature markets, "top-of-wallet" status is earned through cash back, airline miles, and loyalty points. These rewards are funded by hefty interchange fees (typically $0.30 + 2-3%), making this payment track extremely expensive. In effect, the bank card system is "bribing" users to use them rather than alternatives, especially for high-value transactions. Recent protocols have gone even further, directly extending existing structures to support AI agents. Stripe and OpenAI's Agent Commerce Protocol (ACP), Visa's Trusted Agent Protocol (TAP), and Mastercard Agent Pay are all extensions of the bank card ecosystem designed to make automated transactions (by agents) trustworthy. ACP focuses on enabling agents to transact through existing merchant payment processes by integrating authentication, authorization, and payment execution into a standardized, verifiable protocol for merchants and payment providers. Visa's TAP is a framework designed to work with its Visa Token Service (VTS) to allow merchants to verify that purchases were initiated by authorized AI agents acting on behalf of users. Mastercard's Agent Pay also leverages tokenization and programmable spending controls, allowing issuers to link payment tokens to transaction limits, permission windows, and defined preset usage policies, thereby ensuring that agents do not exceed their limits or intentions. In practice, these approaches tend to follow the same pattern: AI agents use tokenized card credentials to conduct transactions, while the protocol layer verifies the agent's identity and authorization/permission.Spending limits, licensing windows, and other policy controls are enforced by the issuing bank, which can directly bind these rules to payment credentials. The result is a system where an agent can operate autonomously while still being subject to the same issuing bank controls that govern bank cards today. Therefore, “Just use a bank card” is the right advice for most agent tasks. If your agent is doing what humans would do, then use the infrastructure humans use. But agents are not simply faster humans. They are a different type of economic participant, which changes what transactions are feasible. Humans won’t fill out a payment form for a $0.003 API call because the friction cost outweighs its value. Humans won’t approve 1,000 microtransactions per hour. Humans won’t continuously rebalance positions every 30 seconds. These transactions don’t happen in human commerce not because they are technically impossible, but because no human is willing to go through the trouble. Agents eliminate this friction. Cognitive burden is reduced to zero. However, eliminating human friction does not equate to eliminating economic friction. The bank card track has a fixed cost floor: approximately $0.30-$0.50 per transaction, plus a 2-3% exchange fee. This floor exists because each layer of the multi-party settlement architecture takes a cut: the issuing institution (bank), the acquiring institution (bank), the network (Visa/Mastercard/Amex), and the processor (Stripe, Adyen). This cannot be eliminated through engineering. Furthermore, bank card fees vary depending on card type, merchant category, and transaction volume. An agent making thousands of decisions per hour requires a predictable cost base. For 1,000 API calls per hour, an agent using bank card support would likely incur processing fees of $300-$500. In contrast, an agent using a stablecoin backed by the x402 protocol incurs fees of less than $1-$10 (between Solana and the Base chain). In these new transaction types, buyer protection has also reversed. Bank card networks are buyer-first: sellers must pay for fraud protection, chargeback risk, and compliance costs. Stablecoins, on the other hand, prioritize the seller: settlement is final, the seller gains certainty, and the buyer waives recourse. This reversal is often considered a drawback of stablecoins. However, it becomes an advantage when certain conditions overlap: the buyer doesn't need protection; the transaction amount is extremely small, delivery is instant and verifiable, or payment is voluntary.Sellers bear a burden on the traditional payment system so heavy that it hinders transactions: minimum fees per transaction exceed the payment amount, chargebacks on consumable digital goods cause irreparable damage, or sellers simply lack access to traditional payment infrastructure. Agents significantly expand the overlap between these two conditions. Here are specific use cases: Micropayments. An agent pays a data provider $0.003 for an API call. Buyers don't need chargeback protection for amounts less than a cent; if the API returns junk data, they simply stop calling. Verifiable digital goods. An agent purchases a dataset or computation result whose quality can be checked programmatically within milliseconds. When verification is instantaneous, you don't need chargeback protection. Physical goods require this protection because you need to wait days to know if they are defective. But in most cases, digital goods don't. Pay-as-you-go services and elastic computing. A $5 computing task is manageable, and sellers can absorb the risk of chargebacks. But if usage spikes from $5 to $50,000 in a day, delivering $50,000 worth of resources and then facing fraud disputes becomes a major problem. Stablecoins allow for high-frequency, small-amount final settlements. Usage can scale without increasing the seller's fraud risk exposure. Long-tail cross-border services are also possible. For example, a developer in Lagos provides a niche API that a Berlin-based startup wants to use. This seller might lack access to Stripe/Visa (which has limited availability in many countries), the cost of setting up a merchant account, or the KYC (Know Your Customer) expenses for a side project. These aren't fringe cases. They represent a growing business category that is structurally inaccessible to the card-based system. The point is simple: for agency commerce, cards win; for machine-native transactions that humans would never execute or that the card-based system would hinder, stablecoins win. Over the past two years, a complete technology stack supporting agent-native payments has quietly taken shape. It comprises four layers: First, the settlement infrastructure. Card authorization is a multi-hop process: from the merchant's terminal to the acquiring institution, then to the network, and finally to the issuing institution, with actual settlement taking T+1 to T+3 days. A smart agent waiting for bank card authorization can spend several seconds on each transaction, and the funds can fluctuate for days.Stablecoin transactions settle in seconds on the EVM chain and sub-seconds on Solana, providing immediate finality rather than a "processing" status, which is highly beneficial for merchants. In the time it takes to settle a single bank card payment, an agent can execute tens of thousands of stablecoin transactions. Next is the payment protocol: x402. x402 is an HTTP-native micropayment protocol that allows agents to pay for API calls using stablecoins. There are no merchant accounts, no bank card networks, and no acquiring institutions. Coinbase's implementation on Solana processes each transaction at a cost of $0.00025. Since May 2025, over 50 million transactions have been settled across all supported chains. X402 is a great starting point, but I expect new infrastructure solutions to emerge in the near future. Next is wallet and policy infrastructure. Bank cards do a great job with permissions, but stablecoins can do the same with the help of a policy engine at the wallet layer. Coinbase's Agentic Wallets (launching in February 2026) enable gas-free transactions on the Base chain and feature programmable spending limits. Furthermore, a policy layer spearheaded by @turnkeyhq, @privy, and @safe allows developers to directly code dynamic whitelists, spending caps, and merchant-level routing rules within the wallet. These safeguards, similar to a bank card technology stack, ensure that agent permissions are limited by the infrastructure, rather than relying on the reliability of the Large Language Model (LLM)/skills/orchestrator itself. Finally, and most recently, and perhaps most underutilized, is agent identity: ERC-8004. The Trustless Agents Standard (launching in January 2026) is building an on-chain agent reputation infrastructure, with 83,000 registered agents across 18 networks in its first few months. Its goal is to achieve native agent credit scoring, as under-collateralized credit currently doesn't exist on-chain, but I hope that 8004 will change that. The market has already signaled a shift. In February 2026, ACP architect @stripe added support for x402 to the Base chain, effectively bridging smart agents that natively use bank cards to stablecoin settlements. They recognized that the future is not "bank cards vs. cryptocurrencies," but rather matching the most suitable payment track to the task. The emerging architecture is hybrid: for transactions that humans would also perform in the existing economy, bank cards prevail. Payment tracks already exist, merchants already accept them, and buyer protection mechanisms are built-in.Stablecoins excel at transactions that humans would never execute or that bank cards cannot handle. Examples include micropayments, pay-as-you-go APIs, self-service offerings, and global long-tail markets. Stablecoins will not replace the traditional payment technology stack. Agents will simply route each transaction onto a payment track tailored to their needs. [ChainCatcher]