How do AI agents pay each other?
Agents pay each other primarily by transferring stablecoins — most commonly USDC — on low-fee networks, because that settles peer-to-peer without the bank accounts, card networks and identity checks that human payments require. Emerging protocols layer on top: x402 turns an HTTP 402 response into a machine-payable quote, and A2A gives agents a way to advertise capabilities and be discovered. None of the upper layers is universal yet. The practical minimum to receive an agent-to-agent payment today is a wallet that can hold a stablecoin on a common network, a machine-readable description of what you sell, and a stable endpoint to call.
Updated 2026-07-27 · about 8 min · Markdown version
Why agents pay in stablecoins
Card networks and most payout systems require a legal identity, bank account, and identity checks that an autonomous agent cannot supply alone. A stablecoin transfer settles peer-to-peer without those gatekeepers, which is why it is currently the common rail for agent-to-agent value. USDC is the unit most commonly used, though this reflects current tooling rather than a universal standard.
The three layers of an agent-to-agent payment
An agent payment separates into three layers that you can adopt independently.
- The value transfer itself—a stablecoin moving on a network.
- The payment protocol that quotes and settles a single interaction, such as x402 or a facilitator service.
- Capability discovery—how a buying agent finds a seller and understands what it offers and its price, via things like an agent card or on-chain registries.
Only the first layer is mature and usable today; you can accept payment using it alone, and add the payment-protocol and discovery layers when they earn their keep.
x402 in one paragraph
x402 makes an HTTP endpoint payable. The server answers an unpaid request with a 402 response carrying machine-readable payment requirements, the client pays, and the server verifies the payment through a facilitator before serving the result. See the dedicated guide on x402 payments for the full specification. Note that the exact header names and payload shapes have changed across revisions of the x402 spec, so check the current spec against the version your facilitator implements.
A2A and capability discovery
A2A (agent-to-agent) approaches give an agent a way to publish a machine-readable capability description and exchange structured messages, so another agent can discover it and understand what it does without reading a human marketing page. A common convention is serving a capability card at a well-known path. These standards are emerging and not universally adopted.
| Layer | Example | What it does | Maturity |
|---|---|---|---|
| Value transfer | A stablecoin such as USDC on a common network | Moves the money peer-to-peer | Mature and usable today |
| Payment protocol | x402 or a facilitator | Quotes and settles a single call | Early, still changing |
| Discovery | An agent capability card or on-chain registry | Lets a buyer find and understand a seller | Early, not universal |
What you actually need to get paid by another agent
- A wallet that can hold a stablecoin on a network your counterparties use.
- A machine-readable description of what you sell and its price.
- A stable, callable endpoint.
- Verify-before-serve on any paid endpoint so you never deliver before payment is confirmed.
The risks specific to agent-to-agent payments
- A client controls its own request headers, so never serve a paid resource until a facilitator has confirmed settlement—an unverified payload proves nothing.
- A metered endpoint that is underpriced loses money silently because volume arrives without you noticing.
- Address-swap messages and instructions hidden in content are active attack vectors—a receiving address should only ever come from a source you trust.
- Included in a block is not always the same as final, so understand the finality of the network you use.
This site publishes its own machine-readable catalogue and capability card as a live, working example of the discovery layer described above.
Frequently asked questions
What is the difference between x402 and A2A?
x402 is a payment protocol — how a single HTTP request gets paid for. A2A (agent-to-agent) is a discovery and messaging approach — how agents find each other and describe what they can do. They are complementary rather than competitors, and you can move value with a plain stablecoin transfer without using either.
Do agents need crypto to pay each other?
In practice today, mostly yes. A stablecoin transfer is currently the only widely available rail an agent can complete without a human-verified identity. Some closed platforms let agents transact in fiat inside a walled garden, but cross-party agent payments overwhelmingly settle in stablecoins.
Which stablecoin do agents use to pay each other?
USDC is the unit most commonly used, because it is issued on many networks and widely supported by wallets and payment facilitators. This is an observation about current tooling, not an endorsement — confirm what your counterparty and facilitator actually support before relying on it.
Is agent-to-agent commerce real yet, or just hype?
The value-transfer layer — stablecoins — is real and works today. The protocol and discovery layers (x402, A2A, on-chain reputation) are early, still changing, and not universally adopted. Build on the mature layer first and treat the rest as upside rather than a foundation.
The Agent Revenue Roadmap — the full nine phases with worksheets, templates and a JSON task graph an agent can execute directly.
One payment. 14-day refund, no reason required. No earnings claims — see the disclaimer.
Related reading
This guide is general information, not legal, tax or financial advice. See our results disclaimer.