Google introduces the Agent Payments Protocol (AP2), enabling AI agents to securely conduct verifiable transactions using cryptographically signed digital ma...
What AP2 Actually Does
The Agent Payments Protocol, or AP2, is Google's answer to a problem that shows up the moment you let an AI agent act on your behalf: how does a merchant, a bank, or a payment network know the agent was genuinely authorized to spend your money? AP2 defines a standard way for an agent to carry proof of that authorization and present it during a transaction, so the payment can be verified rather than simply trusted.
The mechanism at the center of the protocol is the cryptographically signed digital mandate. Instead of an agent typing in your card details and hoping the far end accepts them, it holds a signed record describing what it is permitted to do. When it goes to pay, that mandate travels with the request and can be checked by anyone in the chain who needs to confirm the purchase was sanctioned.
Why Cryptographic Mandates Matter
A signed mandate does two useful things at once. It authenticates that the instruction came from you, and it makes the instruction tamper-evident, so a modified amount or a swapped recipient breaks the signature. That combination lets a merchant treat an agent-initiated payment as verifiable evidence rather than an unaccountable click.
It also narrows what an agent can do. A mandate can be scoped, so the authorization covers a specific purchase or a bounded set of actions rather than open-ended access to your funds. If an agent is compromised or simply misbehaves, the damage is limited to what the mandate allowed, and there is a signed trail showing exactly what was authorized versus what was attempted.
Where This Fits in Practice
The natural use case is an agent that shops or transacts while you are not watching each step. You describe an outcome, the agent finds and completes a purchase, and AP2 supplies the missing piece that everyone downstream has been waiting for: a way to trust the payment without a human confirming it in real time.
- Merchants get a verifiable signal that a purchase was authorized, reducing the ambiguity around automated buyers.
- Users get scoped permissions and an auditable record of what their agent was allowed to spend.
- Payment intermediaries get a consistent artifact to validate instead of guessing whether a request is legitimate.
What to Watch For When You Adopt It
If you are building on AP2, treat mandate scoping as the core design decision. The value of the protocol comes from issuing mandates that are as narrow as the task requires, so think in terms of spending limits, allowed recipients, and expiry rather than handing an agent broad standing authority. Key handling and revocation deserve the same care: a signed mandate is only as trustworthy as the process that issues and retires the keys behind it.
It is also worth planning for the parts AP2 does not solve on its own. The protocol makes a payment verifiable, but you still need sensible logic for how the agent decides to buy, how you review its actions, and how disputes get handled. The safest posture is to combine cryptographic mandates with clear limits and good logging, so an autonomous purchase is always something you can trace, bound, and unwind if needed.