Is crypto necessarily required for payments between AI Agents?

Author: Han Qin, Jarsy CEO

A friend asked whether crypto is necessarily required for payments between AI Agents.

Actually, it’s not always necessary; it’s not that you can’t do it without crypto. However, from a structural perspective, native payments between AI Agents are indeed more easily implemented with crypto. Visa and Mastercard can support some scenarios as well, but there are fundamental limitations.

If AI is just helping humans make payments, traditional payment systems are completely sufficient. But if AI is autonomously settling transactions with other AI, then the structural advantages of crypto are very clear. The fundamental difference between the two lies in whether it’s a human-in-the-loop economy or a machine-native economy.

Can Visa and Mastercard support Agent payments?

Of course, but only in a proxy mode, where AI merely operates payment accounts on behalf of humans. For example, if AI helps you book a flight, it can call your credit card. AI automatically purchasing cloud resources can be linked to a corporate card, and AI SaaS auto-renewals can use Stripe and Visa rails. These scenarios do not require crypto at all.

However, traditional payment systems have structural limitations. First, account systems are tied to identities, requiring risk control, auditing, and the ability to revoke transactions—these are typical human financial system features. Second, traditional payments operate on batch clearing systems, with credit card settlements typically T+1 to T+3, involving many intermediaries and high fees. While acceptable for human consumption, they are not fast enough for a machine economy.

Additionally, traditional payment networks do not support high-frequency small transactions, but the typical payment feature of the AI Agent economy is real-time settlement, microsecond-level payments, and API-triggered automatic payments. Credit cards have minimum fees and transaction costs, making them unsuitable for streaming payments.

Why do people say the Agent Economy is naturally inclined toward crypto? Not because of faith, but because the technological structure truly matches. No identity permission is needed; agents can generate wallets, sign transactions, and trade independently without account approval. This is crucial for a machine economy—otherwise, every AI agent would need to connect to a bank account, which is completely unfeasible. Furthermore, real-time settlement on blockchain is near-instant, irreversible, and without intermediaries, whereas Visa’s network is fundamentally an IOU system.

A straightforward analogy is that Visa and Mastercard are like HTTP plus the banking system for the human internet—suitable for person-to-person and person-to-merchant transactions. Crypto rails are like TCP/IP plus a native settlement layer for the machine internet—designed for AI-to-AI autonomous economies and trustless environments.

The most likely future scenario will be layered collaboration. The top layer involves fiat currency entry points, including Visa, Mastercard, and bank accounts, managed by human funds, with KYC compliance. The middle layer involves stablecoin settlements, such as USDC and tokenized deposits, responsible for fast settlement, API payments, and cross-platform interoperability. The bottom layer is the Agent-to-Agent economy, including wallets, signatures, automatic payments, and machine protocols.

Whether an AI Agent payment scenario requires crypto depends on whether there are ongoing autonomous transactions that do not require human authorization. If yes, then crypto has a significant advantage. If no, then traditional rails are sufficient.

This is also why the AI Agent economy and tokenization are highly coupled. When tokenization turns assets into APIs, and stablecoins turn money into APIs, the next step is for AI Agents to turn decisions into APIs as well. Together, these form a programmable capital markets ecosystem.

Visa is a trust system for human payments, while crypto is a settlement layer for machine trust systems. The two will coexist in a division of labor rather than a replacement or upgrade.