Google’s Agent Payments Protocol (AP2): A New Standard for Trust in Agent-Driven Transactions

Google has unveiled the Agent Payments Protocol (AP2), an open standard developed in collaboration with more than 60 technology and payments companies. AP2 is intended to simplify and secure transactions carried out by AI agents—i.e. software agents acting on behalf of users—by building a reliable protocol that ensures merchants, payment providers, and users can act with confidence, especially in scenarios where a human is not directly clicking “Buy.”

Why AP2 Is Needed

With the rise of AI agents handling commerce—such as automated shopping assistants, task delegators, or bots that act immediately when certain conditions are met—the traditional assumption that a human is always present to initiate a transaction is breaking down. This presents at least three core challenges:

• Authorization: proving that the user has explicitly given the agent the authority to make a particular purchase.
  
• Authenticity: ensuring that what the agent is buying truly reflects the user’s intent (items, price, timing) rather than a misunderstanding or mistake.
  
• Accountability: when something goes wrong—fraud, mistaken purchase, etc.—identifying who is responsible, and having clear evidence to support any dispute or remediation.
  

Google sees AP2 as an extension of earlier protocols such as Model Context Protocol (MCP) and Agent2Agent (A2A), which lay the groundwork for agent interoperability and context sharing. AP2 is meant to handle the payments piece in a unified, payment-method-agnostic way.

How AP2 Works: Mandates & Verifiable Credentials

A key building block of AP2 is the use of Mandates, which are cryptographically signed, tamper-proof digital contracts (or credentials) that serve as verifiable proof of a user’s instructions. These mandates are implemented with what Google refers to as verifiable credentials (VCs).

There are two main workflows:

In both cases, the chain of mandates plus the cryptographic signatures forms a non-repudiable audit trail—i.e., you can later demonstrate who authorized what, that the agent followed the user’s intent, and what was approved for payment. This supports authentication, identity, risk management, fraud detection, and accountability.

Payment Types & Ecosystem Support

AP2 is payment-agnostic; it is designed to work with a wide range of methods, including credit and debit cards, real-time bank transfers, stablecoins, and other crypto-based payments.

More than 60 organizations are already supporting or working with AP2. These include major payment networks, card issuers, fintechs, merchants, and technology companies such as Mastercard, American Express, PayPal, Coinbase, Etsy, and Revolut.

What This Means for Merchants, Users & Developers

Benefits:

• Users can delegate tasks to AI agents more confidently, knowing there is a standard that ensures they remain in control and that their intent is preserved.
  
• Merchants and payment processors gain clarity in verifying whether transactions initiated by agents are properly authorized, reducing risk and potential liability.
  
• Developers can build agents, marketplaces, and experiences using a shared protocol, reducing fragmentation and compatibility issues.
  
• The audit trail created by mandates can aid in fraud prevention and dispute resolution.
  

Challenges & Considerations:

• Adoption: the protocol still needs broad implementation across platforms to become effective.
  
• Regulatory, privacy, and identity verification implications: ensuring that mandates, credentials, and user data are managed in compliance with rules will be crucial.
  
• UX design: making the mandate process smooth for users is essential; too much friction may discourage use.
  
• Edge cases: complex transactions, dynamic pricing, usage-based services, or interactions among multiple agents and merchants may pose implementation complexities.
  

The Road Ahead for AP2 

Google encourages collaboration; the AP2 specification is open and designed to evolve via community input, including standards bodies.

The protocol builds on previous agentic standards (A2A, MCP) to support more advanced commerce models, for example:

• Automated purchase triggers (monitoring stock or price and buying when conditions are met)
  
• Coordinated multi-part purchases (e.g. booking flight + hotel) with consistent mandates across agents/merchants
  
• New business models enabled by agentic commerce (subscription-like tasks, proactive purchasing)
  

As more merchants, payment providers, fintechs, and regulators engage, AP2 could become a foundational layer for “agentic commerce.”

Comparison to Related Protocols

• Model Context Protocol (MCP): focuses on context sharing and preserving user intention across models and agents. AP2 extends it by adding the payments and authorization dimension.
  
• Agent2Agent (A2A) focuses on interoperability among agents. AP2 considers how these agents can securely handle payments, tied back to user intent.
  

Agent Payments Protocol (AP2) seeks to answer fundamental trust questions raised by AI agents acting on our behalf in commerce: when a purchase is made, who authorized it? Is the buying faithful to the user's expectations? If something goes wrong, who is accountable?

By introducing mandates, cryptographic credentials, and a standardized, payment-agnostic framework, Google and its partners are laying the groundwork for safer, more reliable agentic transactions. While real-world rollout and adoption will take time, AP2 could become a significant pillar of how commerce works in an era when “pressing the buy button” may increasingly be done by software rather than by us.