AP2
Agent Payments Protocol

Google's mandate-based payment authorization for AI agents. Three mandate types ensure that every agent purchase has cryptographic proof of human consent -- from browsing intent to final settlement.

Payments Google Open Standard Commerce

What is AP2?

AP2 (Agent Payments Protocol) solves a fundamental problem: how do you let an AI agent spend money on your behalf without giving it a blank check? The answer is mandates -- cryptographically signed authorizations that constrain what an agent can buy, how much it can spend, and when human approval is required.

Released by Google alongside UCP (Universal Commerce Protocol), AP2 handles the payment authorization layer while UCP handles product discovery and checkout. Together, they form the commerce stack for the Internet of Agents.

The three mandate types

AP2 Mandate Flow
User Signs Intent --> IntentMandate --> CartMandate --> PaymentMandate --> Settlement
Each mandate narrows the scope. Intent = "I want to buy software tools." Cart = "These 3 tools, $247 total." Payment = "Charge my card."

IntentMandate

Human declares a goal. "Buy cloud monitoring tools under $500/mo." Agent receives a signed scope to shop within.

CartMandate

Agent selects specific items. Human reviews the cart. Signs approval for exact items and total price.

PaymentMandate

Final authorization. Links to a payment method. Cryptographically bound to the cart -- no last-minute changes.

Why mandates matter

Without AP2, agent payments are binary: either the agent has your credit card and can buy anything, or it can't buy at all. Mandates create a graduated trust model:

AP2 vs x402

AP2 and x402 solve different payment problems:

Enterprise AI needs both: AP2 for procurement workflows, x402 for real-time API consumption.

How MeetLoyd implements AP2

MeetLoyd provides full AP2 mandate support with enterprise-grade governance on top:

  • Full AP2 crypto stack -- ES256 signing, RFC 8785 JSON canonicalization, mandate chain verification. Every mandate is cryptographically bound to its parent.
  • Spending policy enforcement -- Cascading governance policies control per-agent and per-team spending limits. An agent can't exceed its budget even with a valid mandate.
  • Human-in-the-loop -- CartMandate and PaymentMandate transitions trigger approval tasks. Configurable thresholds: auto-approve under $50, require approval above.
  • Audit trail -- Every mandate, approval, and settlement is logged with SOX-grade audit trails. Full chain of custody from intent to payment.
  • Dual-rail -- AP2 for traditional payments, x402 for stablecoin micro-payments. Same governance pipeline, two settlement methods.

See Agentic Payments -->

Learn more

Related terms

Mandate-based agent payments.
That's MeetLoyd.

Agentic Payments Back to Glossary