Catena

The Agentic Commerce Stack: Building the Financial Capabilities for AI Agents

Agentic commerce (a-commerce) is the next evolution of B2B and B2C commerce. It’s an era where AI agents, acting on behalf of businesses and consumers, become primary economic actors. These agents will manage supply chains, procure goods, pay invoices, and handle complex financial tasks with unprecedented speed and scale.

But to unlock this potential, we can't simply plug these agents into our existing financial system. AI agents have dramatically different capabilities and limitations from human actors. They are incredibly fast and can execute millions of transactions without fatigue, but they lack human intuition and are vulnerable to novel forms of fraud and manipulation. A single misconfigured agent could create significant financial damage in seconds.

The New A-Commerce Stack

Realizing the promise of a-commerce requires a new technology and financial services stack, one designed from the ground up to provide the necessary guardrails, security, and services for AI agents to conduct commerce safely and effectively. This stack consists of six layers.

1. Protocols and Payment Rails

At the base of the stack are the protocols for communication and value exchange. This layer must be machine-native. While it will incorporate traditional payment rails like ACH and card networks, it will heavily rely on new, more programmable infrastructure.

Stablecoins will be critical, offering 24/7 settlement, low transaction costs, and programmability. This layer includes new identity and payment protocols such as Agent Commerce Kit (ACK) and x402. It also includes new protocols for how agents interact with each other and with external tools, such as Google’s Agent2Agent Protocol (A2A) and Anthropic’s Model Context Protocol (MCP). This layer will incorporate standards such as decentralized identifiers (DIDs) and verifiable credentials (VCs) from the W3C. These protocols will make it possible for there to be an open, interconnected system to establish identity and trust and to execute payment flows without confining agents to closed networks.

2. Infrastructure

Sitting on top of the protocols is the technical infrastructure—the non-regulated toolkits, frameworks, and systems that make the protocols usable. This is where developers work. It includes:

  • Account and Wallet Infrastructure: Secure systems for agents to hold, manage, and deploy digital assets.

  • Cross-Chain Transfer: Systems for optimizing and managing stablecoin movement across different chains and conversions between multiple stablecoins.

  • APIs and SDKs: Tools that allow developers to integrate their agents into the financial system.

  • Agent Identification: A crucial capability for a-commerce. This is the equivalent of a digital passport for AI agents, allowing them to identify themselves and establish trust with other agents and systems.

  • Rules Enforcement: This is where programmable guardrails are technically enforced. For example, a developer can set rules directly in the infrastructure to limit an agent's spending to $500 per transaction, restrict it to pre-approved vendors, or require human approval for payments over a certain threshold.

  • Payment Rail Infrastructure: These are the technologies to easily integrate and use traditional payment rails such as cards, ACH, and real-time payments (RTP). This layer provides the fundamental building blocks for secure agent operations before any regulated activity takes place.

3. Regulated Banking and Financial Services

This is the layer of trust and compliance, where financial services are provided by regulated entities like banks and Money Services Businesses (MSBs). These services must be re-imagined for an agent-centric world.

  • KYC/KYB becomes KYA: "Know Your Customer" and "Know Your Business" protocols must expand to include "Know Your Agent." Financial institutions will need processes to verify, monitor, and understand the behavior of the AI agents operating on behalf of their clients.

  • Advanced Risk and Fraud Management: Anti-Money Laundering (AML) and Counter-Terrorist Financing (CTF) systems need to be trained to detect anomalous patterns specific to AI agents, which can differ wildly from human patterns.

  • Human-in-the-Loop Controls: This layer ensures that while agents can operate autonomously, there are always clear lines of human accountability and intervention. This is a non-negotiable requirement for managing risk.

  • Deposit Taking and Payment Flows: Taking deposits and handling pay-in, pay-out, and conversion between stablecoins and fiat currencies.

  • Treasury Management: Providing treasury services to generate yield on stablecoin and fiat deposits, including float in payment flows.

  • Credit: Accessing loans and other credit facilities.

  • Auditability and Observability: Actions and transactions executed by AI workflows must be observable and reportable in a compliant, privacy-preserving and human-aligned manner.

Services like deposit taking, card issuing, and custodial stablecoin accounts will all be adapted to serve agents as the primary account users.

4. Fintech Services

Building on the regulated banking core, the fintech services layer packages these complex capabilities into more easily consumable products. This includes Banking as a Service (BaaS) providers who offer agent-ready bank accounts via an API, Payment Service Providers (PSPs) that specialize in orchestrating payments for a-commerce platforms, and embedded finance solutions that allow non-financial companies to integrate these new capabilities directly into their products. This layer acts as an accelerator, allowing businesses to leverage a-commerce without needing to build a direct, deep integration with a bank.

5. Vertical and Horizontal Business Solutions

This is where the technology meets specific, real-world business needs. Here, the underlying stack is used to build proprietary corporate systems or commercial software products.

  • Horizontal Solutions: These address common business functions, such as agent-driven accounts payable (AP) and receivable (AR) management, intelligent payroll, and automated international remittances.

  • Vertical Solutions: These are industry-specific applications. For example, a logistics company might use an agent-based system for supply chain management that automatically orders materials, tracks shipments, and pays suppliers upon delivery confirmation.

  • Consumer Solutions: These might include agentic services for shopping, organizing travel, or investment and budget management.

  • Agent Solutions: Services that give agents access to commercial capabilities such as proprietary data sets, analytical tools, or tools for executing real-world actions.

This is the layer where most businesses will directly experience the benefits of a-commerce through increased efficiency and automation.

6. End Users

The final layer is composed of the businesses and consumers who use the solutions. A manufacturing company's procurement department uses a vertical solution to reduce costs, or a small business owner uses a horizontal accounting platform to automate invoicing and payments. For consumers, this could mean a personal finance agent that negotiates bills, finds the best insurance rates, and executes trades based on pre-defined goals. The complexity of the underlying stack is abstracted away, and users simply benefit from the outcome: smarter, faster, and more efficient commerce.

The Path Forward

Building this stack will be an evolutionary process. It will not be a "rip and replace" of the existing financial system. Instead, it will be a collaborative effort. Existing players --- banks, payment networks, and established fintechs—will move to adapt their offerings for the agent economy. They bring regulatory expertise, trust, and scale. At the same time, new entrants will emerge, driving disruptive innovation at every layer in the stack.

As recognized by analysts at firms like Gartner, who have highlighted the importance of AI Trust, Risk, and Security Management (AI TRiSM), the focus must be on creating a robust and secure environment for AI to operate. The success of a-commerce hinges on our ability to build a financial system where AI agents can act as powerful economic engines while being safely tethered to human oversight and sound risk management principles. This six-layer stack provides the pragmatic roadmap to get there.

Where We'sre Building

At Catena Labs, we’re working on advancing the work at several layers in this stack. We’re developing protocols and pattern languages like the open source Agent Commerce Kit (ACK) and contributing to other protocols we believe will be critical at the protocol layer. We’re also building infrastructure that will be required for developers integrating and using these lower-level protocols by enabling easy adoption of agent identity, stablecoin wallets, and rules. Our main focus is working at the banking and financial services layer to provide the regulated, reliable, and trusted financial services that will be required for a-commerce to thrive.