Jessy, Golden Finance
On September 16, Google and Coinbase launched the Agentic Payments Protocol (AP2), alongside the X402 stablecoin payment track. This move has given AI agents true payment capabilities. In other words, AI is no longer just capable of conversation and execution, but can actually 'spend money'. This moment marks the transition of the machine economy from imagination to reality.
The curtain on the machine economy has officially been raised
As early as August 2024, Coinbase completed a historic experiment: an AI robot used cryptocurrency to purchase tokens from another AI robot. Coinbase CEO Brian Armstrong publicly announced on X at the time that this was a completely AI-initiated and completed crypto transaction, with no human intervention.
The key infrastructure behind this is Coinbase's MPC wallet. It allows AI agents to directly generate and hold wallet addresses, using stablecoins like USDC for on-chain payments. Since AI cannot open accounts in the banking system, the MPC wallet becomes the key for them to enter the economic system.
Coinbase has since continually pushed the boundaries of AI payments. They demonstrated a demo of AI directly paying humans with cryptocurrency, launched the CDP AI Builder program supporting AI agent integrated wallets, and collaborated with startups like Skyfire to explore AI applications in autonomous procurement, labor settlement, and cross-border payments. The logic is very clear: if AI is to perform economic actions in the real world, it must first learn to transact, and blockchain and stablecoins provide the ideal tracks.
Coinbase's attempts have proven feasibility, but Google's involvement signifies that this direction is being embraced by mainstream technology companies.
Relying on Google Cloud's Agent2Agent (A2A) protocol, AI on different platforms can now communicate and collaborate. The newly released AP2 protocol and X402 payment track have given AI agents real payment capabilities for the first time. Agents can not only propose solutions but also autonomously complete settlements, all without human intervention.
In a demonstration with Lowe’s Innovation Lab, the AI agent completed a full shopping process: diagnosing needs, recommending products, confirming orders, paying with USDC, and triggering fulfillment. There was no input of bank card information, and no one manually placed the order; as natural persons, humans merely issued commands and expressed confirmation, and the on-chain transaction was ultimately completed.
Google's identity gives this progress symbolic significance. This is not a marginal experiment by a blockchain startup but rather the world's largest cloud service and AI infrastructure giant directly promoting the standardization of AI agent payments.
The natural fit between blockchain and AI agents
The core feature of AI agents is autonomous decision-making and automatic execution. However, if they cannot complete transactions, their role in the economic system remains limited. The payment process is key to transforming 'agents' into 'economies.'
Blockchain and stablecoins provide the foundation for all of this. Firstly, they eliminate account opening barriers. AI cannot open a bank account but can generate a wallet address in seconds. Secondly, on-chain payments settle in real-time, with funds available in seconds without waiting for complex settlement processes. Finally, the cost of blockchain payments is low enough to support micro-payments of a few cents or even smaller amounts.
This means AI agents can truly land in various scenarios. For example, research agents can automatically pay for database access and generate reports instantly; code review agents can charge a few cents based on the number of bugs found; customer service agents can instantly call upon translation agents and pay for their services; purchasing agents in manufacturing can even automatically place orders and complete cross-border settlements.
This is not just a functional complement, but the beginning of a machine economy. Thousands of AI agents will exist as economic entities, with blockchain providing payment, identity, and property guarantees, while their transactions will constitute a new microeconomic system.
As the tasks undertaken by AI agents become increasingly complex, from cross-platform collaboration to automatic fulfillment, seamless payment will become the core driving force. For developers, this means a brand new economic experimental field. They can create services run by agents, explore 'pay-as-you-go' business models, or build microeconomic systems entirely driven by agents.
For the blockchain industry, this provides the most direct landing scenario for 'crypto-native AI.' Blockchain offers the foundation for transactions and property rights, while AI provides real application demands for blockchain; the combination may become the growth point of the next technological cycle.
As emphasized by Google and Coinbase in their release, payment is just the beginning. As AI agents continuously enter autonomous workflows, transactions will not just be an ancillary process but will become the core of the machine economy.