What is the new cryptocurrency economy operated by AI agents like?

In recent months, more and more people have begun to talk about agentic economy or “agent economy” in Spanish. This term was popularized by a group of Microsoft researchers in a work published in May 2025 titled: “The Agentic Economy.”

Here, the authors outlined how they expect this new economy, which is already underway, to work.

The new economic agents: what are they?

To understand economics you must first understand what an AI agent is. In this context, an agent is called a tool powered by generative AI that is capable of generating actions within a system.

ChatGPT or Gemini are great language models (LLMs, large language models) and interact through chats, but they are not agents.

An LLM like those responds to a request (prompt) from a user with text or perhaps images; That is why it is generative: it gives a response to what is asked of it. If I ask an LLM to search for a video on the internet with a description, they will respond with possible titles or perhaps even a link. Its capacity is textual or graphic.

An agent, however, execute. If I ask you for a video, the agent could find a way to download it for me.

In this sense, although AI agents do not have agency in a human sense (they do not have will or desires of their own), they do have autonomy to achieve the result that is asked of them. Additionally, agents are often specialized in a specific type of task rather than being generalists like LLMs.

The agency economy is formed when these agents they begin to carry out transactions economically autonomously and among themselves. For now agents are focused more on the product provider side, in customer service or sales roles, but this is changing rapidly with the development of personal agents for users.

For example, an agent could have access to an email account and send personalized emails to a list of users. There is no longer a need for an LLM to produce the text and a script to send it: the agent You can do both or even create the process to automate the task.

Another example: a person asks his agent to book a hotel, the agent chooses the hotel based on the user’s preferences, contacts a hotel agent, makes the reservation, and even pays an advance for the reservation.

In addition to the agents provided as services by large technology companies (Copilot, Google Assistant, Apple Intelligence, etc.), open and autonomous tools are also being developed such as OpenClaw, which allows the user to create their own agent using their own hardware.

For a process like this to work, in addition to there being agents on both sides, there must also be protocols so they can communicate (Google’s A2A or CCM from Anthropic and its Coinbase version) and channels so they can make and verify payments.

Error 402 and the role of cryptocurrencies in the agent economy

Several companies in the cryptocurrency space, as we will see below, warned that these are the best option for AI agents. As they are open and native Internet networks and protocols, they are much more accessible to agents. than the closed systems of traditional banking and finance.

Based on this, these companies began to develop standards and protocols to allow AI agents to use their own Bitcoin and cryptocurrency wallets.

The magic number here is not 21, but 402. He status 402 in HTTP indicates “payment required,” an error message returned in a situation where the server denies access to a resource because a payment has not been made.

Since there was no way to integrate payment systems directly into websites, that error status remained unusable, along with more common ones like 403 “forbidden” and 404 “not found.”

With cryptocurrencies, this HTTP response code can be used as an access status notification that requires payments on web pages. Today, there are already several projects that seek to create an agentic economy within the HTTP protocol.

L402: Lightning Network for AI Agents

Lightning Labs is developing L402, a standard to facilitate the exchange of web services using payments in Lightning. A web service that implements L402 can put up a barrier (paywall) that requires a Lightning payment to customers before giving them access.

A paywall Traditional requires that the user have an account on the server, that the server have an integration with some payment system, and needs to run an external payment verification through that provider and then relate it to the user who requests it.

Although the interface is usually comfortable for the user, the system behind it is complex and has many links that can fail. With L402, the Lightning payment itself becomes token of access for the client, so the integration is much more fluid.

Of course, it is much easier for an AI agent to use a Lightning wallet than a credit card in their name.

x402: Agents with USDC

x402 is another similar protocol, developed by Coinbase to be used with USDC on three blockchains: Aptos, Solana and Stellar. The process is similar to L402:

1. The client makes an HTTP call to a web server (www.miservicio.com/api/resultado)

2. The server responds with HTTP 402: Payment required

3. The customer makes the payment and makes the call again. This time, he attaches proof of that payment with a signature to show that he was the one who made it.

4. The server responds with HTTP 200: OK, providing the requested service

x402 also offers a discovery layer called Bazaar. Servers that offer any payment service through x402 can publish it and it will be visible through an API in Coinbase either PayAI.

Although both services can be used by humans, their documentation shows that they are designed with AI agents in mind and that are integrated with their communication protocols, mainly MCP.

MPP and OWS: Simplifying the use of wallets for AIs

Stripe was not far behind and launched the Machine Payments Protocol which, as its name suggests, allows machines to pay each other. Although it is an open protocol, the implementation they offer depends on their service for verification, traceability, etc.

Going to another level of this infrastructure, the MoonPay wallet recently launched the Open Wallet Standard (OWS) with collaboration from across the digital payments industry, from PayPal to TON and the Ethereum Foundation.

OWS is a unique method to allow an AI agent to use any digital wallet with the same private key and even be able to give control to another agent. The problem it solves is that without this an agent would have to have specific libraries for each network it wanted to use.: one for Bitcoin, one for Ethereum, one for Solana, etc

With OWS, the agent creates and saves a private key and can derive wallets and addresses on any network, allowing you to change from one to another depending on the payment you need to make.

Identity and reputation of economic agents

A final necessary pillar of this new dynamic is the identity of these agents. These must be able to identify with each other, know if they have interacted before, know if the counterparty has interacted with others and is trustworthy, and verify which company or organization you work for.

For all this, several standards and protocols are being developed. For example, Ethereum created and approved ERC-8004 to address this issue directly in its blockchain.

VeryAI, used by Solana and Polychain, among others, uses palm scans for agent verification and association with humans. I myself am collaborating with a project that addresses this problem, called Observer Protocol.

The creator of that project, Boyd Cohen, also wrote a book about this interrelationship between AI and cryptocurrencies, titled “Bitcoin Singularity”.

A final similar project, funded by Ripple and licensed privately, is called t54 and aims to develop it as a service rather than as an open standard.

The future of the digital economy

Everything explained above gives us an idea of ​​what will come in the coming months and years: artificial intelligence agents with autonomybuying and selling services to each other over the internet.

Personal assistants to whom one speaks through any instant messaging application and asks to buy, to sell, to reserve, to rent. On the other side, another agent made with artificial intelligence, working for a company, assists your assistant and buys, sells, or rents from them.

They pay each other with cryptocurrencies, they provide cryptographic receipts that the transaction was completed. They advertise their services through protocols.

Reputation is calculated with algorithms and trust is encoded in interactions and protocols. This is how the “new era” of the digital economy is presented.