From Visa to Coinbase, here’s how big companies will approach AI payment infrastructure in 2025

Penulis asli: cookies

Terjemahan asli: TechFlow

Hey, welcome back! This is part 2 of this series. If you’re wondering where part 1 is, you can find it here. I highly recommend reading part 1 first so you understand the necessity of blockchain in autonomous capital before diving into the companies working on autonomous capital infrastructure.

The Age of Agentic Capital | Part 1 So…did you watch Part 1? If not, go watch it now – just kidding. If you didn’t watch Part 1, here’s one thing you must know before reading this:

Autonomous capital: refers to artificial intelligence systems that have the ability to independently hold, manage and deploy financial resources to achieve specific goals without human intervention. In this context, agents refer to autonomous economic entities that have their own economic behavior capabilities.

perkenalan

This article will take you through a range of topics, starting with an understanding of the existing financial system. This lays the foundation for understanding how to integrate agents into the system to enable autonomous commerce (i.e. agents pay you). Finally, we will provide a comprehensive overview of the companies that are building the autonomous capital infrastructure to enable autonomous commerce.

Traditional financial system

Often referred to as the “four-party model,” this is a common approach to payment processing:

The Four-Party Model of Payment Processing

1. The user pays using a credit card.

2. The card information is passed to the acquirer.

3. The acquirer passes the card information to the payment network.

4. The payment network asks the card issuer to authorize the payment – the card issuer verifies the card details and checks for sufficient funds.

If you want to know more about how payment settlement is implemented, you can refer to this article.

Further analysis:

• Issuer: An issuer is a financial institution that issues credit cards to customers. An issuer can be a bank, credit union, or other financial institution that chooses to provide credit to customers. When a customer disputes a transaction on their credit card, the issuer is responsible for deciding whether to grant a chargeback. Some of the largest issuers include Chase and Bank of America.

• Acquirer: An acquirer is a bank or financial institution that collects payments from card issuers on behalf of merchants. The acquirer is responsible for ensuring that customer payments are completed by passing information to the payment network. When a chargeback occurs, the acquirer is responsible for returning the customers funds (ultimately deducted from the merchant).

• Payment Network: A payment network processes card transactions by connecting customers, merchants, acquirers, and card issuers. Common payment networks include Visa and Mastercard.

If you want to learn more about the various parties involved in the payment process, you can refer to this article .

The rise of autonomous business

Now, we’re passing the baton to “agents” – they’re going to be the new shopaholics.

Agentic Commerce: Simplifies the shopping experience through an agent-led payment process, which also includes search and recommendation capabilities.

Autonomous commerce is not the only form of autonomous capital, but in this article we use it as an example of why agents need to have the ability to access funds autonomously.

The popularity of autonomous business is evident:

sumber

Sumber It is not the purpose of this article to delve into why these tech giants are trying to establish a foothold in the autonomous commerce space. At a macro level, here are some of the advantages that companies can gain by becoming the consumer interface layer for autonomous commerce:

• Economic benefits: Every time a user completes a payment on the platform, the platform can earn fees.

• Data flywheel effect: Platforms can gain valuable insights from agents’ purchasing patterns and analyze this data to recommend complementary products to users — imagine YouTube’s algorithm always pushing the videos you want to watch.

• Network Effect: When a platform becomes a source of product discovery, it attracts more merchants to list products on its platform.

You can read this great article by one of my team members, Evan, which details how business models and advertising models will evolve as the business paradigm shifts. I personally think that the topic of agency and advertising monetization is an underrated area, and this article does a great job of explaining the nuances of this new digital era.

Autonomous capital infrastructure for traditional companies

Youve seen big tech companies try to take advantage of the autonomous business world. So how exactly are they doing this?

This section will detail the autonomous capital infrastructure (TechFlow Note: more commonly known as AI payment infrastructure at this stage) built by several traditional financial institutions (including Visa, Stripe, PayPal, Coinbase and Mastercard).

sumber

Visa | @Visa

On April 30, 2025, Visa announced the launch of the Visa Intelligent Commerce framework, which aims to enable artificial intelligence to shop and pay autonomously. The framework is designed to treat agents as end consumers.

Architecture and functional highlights:

• Payment Passkeys: used for identity authentication during agent checkout. This authentication increases merchants’ confidence in accepting payments and promotes the popularity of autonomous commerce.

• Access Controls: Payment keys enable users (proxy owners) to specify parameters including spending limits, approved merchant categories, etc. These capabilities are implemented through APIs and SDKs designed by Visa.

• Dispute Resolution: Specially used to deal with problems that may arise during the agency payment process.

Workflow:

1. The user enters the demand for a product in the Visa Agentic Wallet.

2. The agent starts searching for products.

3. After receiving the search results, the user authorizes the agent to complete the payment using the payment key.

User Authorization

You can check out the demo video .

Keuntungan:

• Trust: By authorizing transactions, users can rest assured that the agent will not lose control and make inappropriate spending.

• Network Effect: 99.9% of merchants already accept Visa, which means users can buy almost anything.

• Personalization: By studying the transaction patterns of users, agents can understand the preferences of users and make better decisions over time, such as choosing a hotel that fits their budget.

• Strong partner lineup: including leading language model users such as OpenAI, Perplexity, Microsoft, Anthropic, Mistral AI, etc.

For more information on Visa’s autonomous payments architecture, check out this article .

Stripe | @stripe

Back in November 2024, Stripe launched the Stripe Agent Alatkit, allowing businesses to integrate payment capabilities into their autonomous workflows.

Architecture and functional highlights:

• Disposable virtual card: Designed for agent consumption and implemented through a simple language model function call.

• Access control: Parameters for agent consumption, such as budget and merchant selection, can be defined through the order intent API.

• Native billing support: Stripe monitors payments and charges businesses based on usage.

Workflow: Here is an example of an agent helping a user book a flight:

1. Agent builders conduct Know Your Customer (KYB) with Stripe to obtain API keys, giving their agents autonomous fund management capabilities.

2. The user prompts the agent to “find me a flight to Rome for less than $1,000.”

3. The agent searches and finds a flight for $800.

4. Use Stripe Issuing to generate a one-time virtual card that can only be used to spend $800.

5. The agent generates a payment link, which is authorized by the user.

6. Once the payment is completed, Stripe notifies the agent and updates the status, and the card is closed.

For a visual representation of the workflow, check out this rekaman video , and for more detailed information on the architecture, check out Stripe’s developer documentation.

Keuntungan:

• Easy access: Agents can complete payments through traditional financial payment methods, including credit cards, bank transfers, Apple Pay, etc.

• Widely integrated: The toolkit has been integrated with widely used platforms, including OpenAI’s Agent SDK, CrewAI, LangChain, and Vercel’s AI SDK.

PayPal | @PayPal

On April 14, 2025, PayPal launched the PayPal Agent Toolkit, which supports building autonomous workflows that can handle financial operations.

Architecture and functional highlights:

• Full process support: Full process support is achieved through PayPal account association, kripto wallets, and the emerging payment key checkout process.

• Core business functions: Agents can access payment, invoicing, dispute resolution, shipment tracking, product catalogs, subscriptions, reporting and more.

Workflow:

1. Users search for products through Perplexity.

2. After finding the product, users can check out quickly through PayPal or Venmo (identity authentication required).

For more information about Google Clouds Proxy-to-Proxy (A2A) protocol, see halaman ini .

Merchant application scenarios:

• Order management and shipment tracking: Agents can intelligently handle order status and shipment information.

• Smart invoice processing: Agents can generate invoices based on predefined templates or dynamic parameters, send them to customers, track payment status, and provide reminders for overdue payments.

• Simplify subscription management: AI agents can manage the entire subscription lifecycle, including creating new products, subscription plans, and processing recurring payments through payment methods supported by PayPal.

kekurangan:

• Network effect: Only about 72.5% of retail websites accept PayPal, which is slightly lower than the 99.9% merchant acceptance rate of other payment networks.

Coinbase | @coinbase

On May 7, 2025, Coinbase announced the launch of the x402 protocol, a payment protocol that allows stablecoin payments directly over HTTP.

Architecture and functional highlights:

• Utilize the original HTTP status code “402 Payment Required” to embed stablecoin payments into web page interactions.

• x 402 allows developers and AI agents to pay for APIs, services, and software directly using stablecoins over HTTP.

Workflow:

• The proxy requests the required resource from an HTTP server that supports x402 (such as GET /api).

• The server responds with a 402 Payment Required status code and provides payment details (e.g. price, acceptable tokens).

• The client sends a signed payment packet using a supported token (such as USDC) via standard HTTP headers.

• The client resends the request and includes the encoded payment packet in the X-PAYMENT header.

• The payment facilitator (such as Coinbase x 402 Facilitator service) verifies and completes the on-chain payment and fulfills the request.

• The server returns the requested data to the client and confirms the success of the transaction in the X-PAYMENT-RESPONSE header.

Keuntungan:

Metered services: Users can make small payments instead of paying a one-time high fee

Mastercard | @Mastercard

Mastercard has launched Agent Pay, an agent payments initiative that seamlessly integrates payment capabilities into its conversational AI platform.

Architecture and Functionality

1. Agentic TokenS:

   1. A system for registering and verifying agents, equivalent to a KYC (Know Your Customer) system, used to authenticate agent identities.

   2. Tokenization: Protect information security by replacing the card number with an alternative number.

   3. When you use the card, a one-time code (encrypted information) is generated to authenticate the transaction.

   4. Learn more about tokenization in detail .

2. Payment Passkeys:

   1. Identity confirmation through equipment (such as biometrics).

3. Access Controls:

   1. Consumers can define what the agent allows them to purchase.

Keuntungan

• Transparansi: Proxy tokens make it possible to identify and track transactions performed by the proxy.

• Partner: Work with acquirers and checkout providers like Braintree and checkout.com.

To learn more about Mastercard partners, check out this article .

Blockchain Autonomous Capital Infrastructure

The next section will delve into companies that focus on building autonomous payments using stablecoins. Before that, let’s first understand the two key players in autonomous payments:

Agent Builder: A developer who develops an agent that provides a specific service and receives payment for it.

Agent User: A user can deposit a certain amount of funds into an agent for use, and the agent will then make the payment.

Skyfire | @trySkyfire

Skyfire is committed to becoming a payment network similar to Visa, creating a global, instant, and secure payment network specifically for AI agents, allowing agents to make and receive payments autonomously. This will enable agents to access resources such as MCP servers, APIs, large language models (LLMs), and datasets.

Architecture highlights:

• Stablecoin: supports programmatic payments.

• Unique Identifier: Creates a verified identity for agents to open an account and start receiving and initiating payments.

• Trust Badges: Various vendors offer “Blue Tick Certification” (similar to a blue checkmark) to agents, allowing businesses on the Skyfire platform to trust and interact with the agent (whether buying or selling).

Features:

• Agent-to-Agent Payments: Supports instant payments between agents.

• Autonomous Payments: Agents can complete payments without human involvement and can receive payments without a bank account.

• Agent Funding: Agents can obtain funding through a variety of methods, including debit cards, credit cards, wire transfers, and stablecoins.

• Access Control: You can set consumption limits for each agent.

• Agent History: Provides a verifiable record of agent activity. For agent builders, this enables them to track demand for their agent services.

Skyfires proxy tracking interface

Workflow Here is an example of an agent trying to purchase data access:

• The agent initiates a service request by calling the data providers API.

• The service verifies the identity of the proxy.

• After authentication, the proxy receives the data it requested.

• Once the transaction is complete, the payment is released to the data provider.

Keuntungan

• Pasar expansion: Companies can sell services not only to human users, but also to agents (an under-exploited market).

  This pattern is particularly striking in long-tail markets.

• Traditionally, selling long-tail services to enterprises has been difficult because the small number of customers makes it difficult to justify the costs of setting up a dedicated business to sell a small number of products, including business development (BD), user acquisition, etc.

• With a universal pasar, you can simply upload content such as long-tail datasets, and agents will discover these resources themselves based on metadata, eliminating business development and advertising expenses.

Industry Application and Impact

• Pricing culture : Real-time contextual data intelligence on proprietary data sets is being purchased and consumed by agencies.

• Denso : Supporting agents in sourcing required materials in the global automotive parts manufacturing industry.

Financing News

• Skyfire received $9.5 million in financing supported by Coinbase Ventures, a16z CSX, Circle, etc.

Payman | @PaymanAI

Payman is a platform where AI agents can pay humans for professional tasks completed (think of it as AI outsourcing tasks to freelancers).

Architecture

• Human Marketplace:

  Payman is developing a database of vetted skilled workers to ensure high-quality task completion. The marketplace provides AI agents with resources for human tasks.

• Verification Agents:

  A set of specialized agents that verify that work completed by humans meets task requirements.

• KYC Verification:

  Individuals who wish to join the Payman worker database are required to undergo KYC (Know Your Customer) verification to ensure that the agent is paying for legitimate services.

• Wallet system:

  USD Wallets: Stored in Stripe.

  USDC Wallets: Hosted by Bridge.

• Payment Networks:

  Payman uses Skyfire to implement agent-to-human payments.

Feature Highlights

• Payment Methods: Agents can be paid in a variety of ways, including fiat and cryptocurrencies.

• Programmable policy enforcement: Each transaction is governed by user-defined rules, such as a specific amount to be paid for a task.

• Payment Tracking: Each transaction has a reference ID and real-time status for seamless tracking.

• SOC 2 Compliance: A framework for assessing and validating information security practices for customer data, Payman uses data masking and encryption to protect sensitive information.

Workflow

Put yourself in the shoes of someone who wants to build an agent that will ultimately be responsible for finding resources for task completion.

For users who wish to set up a proxy:

• Wallet setup: The user sets up a wallet for the agent and adds funds.

• Task creation: The agent creates tasks based on user prompts and publishes them to the task marketplace for human workers to discover.

• Policy setting: Set spending limits, approval rules, etc.

• Task Approval: Users receive a notification on the dashboard and can approve or reject the agent request.

For human workers:

• Add a wallet (crypto wallet or bank account) to the platform.

• View Payman Platform Demo

Skenario Aplikasi

• Collect user feedback to improve product development.

• Human Resources: Pay candidates for task completion.

• Software Engineering: Pay experts to do code reviews.

Pembiayaan

• Pre-seed raised $3 million in funding, supported by Visa, Coinbase Ventures, and others.

My thoughts

• How sophisticated is the verification agent? For tasks where the outcome is clear, it is relatively easy to verify that the work was done well. However, for tasks where the outcome is less clear, what kind of agent will be used to verify?

• Payman currently requires users to manually create tasks. So, in the future, is it possible to give agents autonomy to judge which services need to be purchased based on general prompts? This may be achieved through small language models (SLMs) that focus on specific fields, such as in the automotive industry, the restaurant industry, etc.

• Overall, this idea is very interesting. It injects some human qualities into algorithmically programmed agents. In the age of AI, this approach helps preserve human aesthetics and diverse perception. For example, a designers evaluation of the agents completion of a task can add more value and depth to the result.

Catena Labs | @catena_labs

Catena Labs is building the first AI-native financial institution, a regulated entity that facilitates collaboration between agents and humans. Catena Labs designs with the vision that agents will become powerful economic actors, building infrastructure including agent identification and an AI-specific risk management framework.

Architecture – Agent Commerce Kit (ACK)

ACK is designed to enable interoperable financial interactions involving agents and consists of two complementary protocols:

• ACK-ID: Verifiable proxy identity

  Verifiable Agent Ownership: Agents are cryptographically linked to their organizational owners (humans).

  Security authentication: Used by agents to prove their identity to counterparties and systems with which they interact.

  Privacy-preserving authentication: only necessary identity information is disclosed.

  Built on decentralized identifiers (DIDs) and verifiable credentials (VCs).

  Learn more about ACK-ID

• ACK-PAY: Proxy Native Payment

  To learn more about ACK-PAY, please see this page .

  Providing infrastructure that enables agents to initiate payments and conduct financial transactions.

  Standardize payment processes: Communicate payment requirements from service providers to agents.

  Flexible Settlement: Agents can settle transactions through a variety of payment channels, including traditional finance and cryptocurrencies.

  Verifiable Receipts: Provide cryptographic proof of payment via secure credentials.

  Human oversight: Certain areas of the workflow still require human approval.

Workflow Here is an example of an agent attempting to purchase proprietary financial data:

• Request: The agent sends a request to the organization M (M stands for Money) to purchase financial data.

• Authentication: Organization M uses ACK-ID to verify the agents identity (ensuring that it is not malicious) and its access rights (whether it has the right to access the data).

• Payment:

  Organization M sends a standardized ACK-Pay payment request payload to the proxy.

  Agents select the payment method (fiat/crypto) based on their preference.

  This process may require human approval.

• Receipt:

  After the payment is confirmed, a verifiable ACK receipt is generated (as a verifiable credential).

  The receipt is delivered to the agent, who then gets the required data.

View more detailed workflow

Financing News

• Catena Labs received $18 million in financing supported by a16z Crypto, Circle Ventures, Coinbase Ventures, and others.

My opinion is that Catena Labs seems to be building a comprehensive infrastructure solution for “proxy capital”. Its development plan includes enhancing the proxy identity mechanism, protocol interoperability, compliance and risk monitoring tools, and establishing a proxy reputation system.

Nevermined | @Nevermined_io

Nevermined is building a payments platform designed specifically for AI agents, enabling them to initiate and receive payments.

Workflow

Proxy setup phase

• Agent Registration: AI developers register their agents and payment plans through the Nevermined app.

• Payment Plans: A program that records how much a developer wishes to be charged for their agency – Nevermined charges a 1% processing fee when a plan is sold.

• Planned Creation: Nevermined associates decentralized identifiers (DIDs) with developers’ wallets.

• Discovery Settings: Nevermined creates a widget with plan information for display on its marketplace.

User access process

• Purchase: Users order a payment plan through the Nevermined Marketplace.

  If a user chooses to pay the agent via stablecoin, the funds will be locked in a smart contract before the user can claim their points.

• Distribution: Users receive points corresponding to their purchased plans.

• Consumption: Users make requests to the AI agent through the Nevermined application.

• Access Control: There are two types of points that determine how the agent can be used:

  Time-based: Users are granted access for a fixed period of time, such as 1 day or 1 month.

  Request-based: Users redeem points based on requests, divided into:

  Fixed: Each request consumes a fixed number of credits.

  Dynamic type: The point consumption changes dynamically according to the complexity of the request.

To learn more about payment types, check out the detailed article about Nevermined Payments

Keuntungan

• Discovery Engine: Nevermined becomes an agent discovery platform by storing the metadata provided by agents when they register.

  If you are a developer and want to monetize your agents, click Di Sini . If you are a user and want to hire agents, click Di Sini .

Fewsats | @fewsats

Fewsats enables agents to pay for services without relying on external payment services.

Architecture and Functionality

• L 402: When the server responds with an HTTP 402 Payment Required status, it includes machine-readable payment terms for the agent to complete the payment.

• Authentication: The identity of the agent is authenticated through a cryptographic proof protocol.

• Access Control: Define transaction limits and approval workflows.

View Demo

There is little information about Fewsats at the moment, but there are reports that it may integrate the Bitcoin Lightning Network to support low-cost, near-instant payments for agents.

Kesimpulan

In this article, I’ve covered the following companies active in the AI payments infrastructure space: Visa, Stripe, PayPal, Coinbase, Mastercard, Skyfire, Payman, Catena Labs, Nevermined, and Fewsats (to be added in the future with information on Nekuda, Protegee, and Brahma Finance). This is undoubtedly a rapidly evolving and emerging space, as evidenced by the innovations of traditional payment companies and the investments made by payment networks.

Over the next few years, key infrastructure questions will need to be answered, including:

• Proxy Identity Verification: How will KYA (Know Your Proxy) be achieved at scale?

• Access control: How granular will it be? How much autonomy can be delegated to agents? Does this autonomy increase with increasing “good behavior”?

However, some areas remain questionable:

• Source of funds: Currently, most infrastructures are verifying the identity of agents to ensure that they are not malicious and have funds. Once agents start to engage in larger transactions, will this raise concerns about money laundering?

From reading about the architectures and products of various companies, I believe that the companies that dominate this space will be those that can develop applications that allow retail and enterprise customers to seamlessly access the services they need through the proxy. Will this become a business development game to see which platform can attract the most partners/providers – perhaps.

In my opinion, the data generated by these agent transactions will be extremely useful and can generate a data flywheel. For example, in the case of purchasing agents, their purchases can be optimized by learning from their transactions and interactions with other agents. And for agents providing specific services, they can observe the purchasing behavior of other agents and potentially create new services to capture a larger market.

This is the second part of the series on proxy capital. In the third part, we will continue to explore its practical application scenarios.

Penafian This article is based on company documents and articles. If there are any inaccuracies, or aspects you disagree with, please leave a comment and I will be happy to discuss further.

Tautan asli

This article is sourced from the internet: From Visa to Coinbase, here’s how big companies will approach AI payment infrastructure in 2025

Related: Odaily Planet Daily Trump Dinner Special AMA Interview with BitMart Founder Sheldon

On May 21, Odaily and BitMart jointly held the Space event. BitMart Founder Sheldon ( @sheldonbitmart ) was a guest in the live broadcast room and exclusively revealed the inside story of the Trump dinner. Sheldon said that the specifications and significance of this dinner were significantly higher than the Trump 100-person charity dinner he had participated in before. He pointed out: This dinner is not only an important historical opportunity, but also an important symbol of the Web3 industry breaking through traditional barriers and integrating into mainstream society, promoting blockchain technology towards a wider global development. The following is the content of Space: Hello, welcome to the Space of Planet Daily. I am Qian Xiaofeng, deputy editor-in-chief of Odaily Planet Daily. In recent times, the most lively market is…