Visa and Cloudflare launch trusted agent protocol

The world of e-commerce is on the cusp of a significant shift, moving from direct human interaction to a machine-mediated process where AI agents shop and transact on our behalf. This transformation, dubbed agentic commerce, presents immense opportunities but also novel risks, prompting key industry players to act quickly to establish foundational safeguards.

The immediate need for security in this evolving landscape is highlighted by the introduction of the Trusted Agent Protocol by Visa and Cloudflare. This new framework is designed to secure transactions as AI agents increasingly shop and pay for consumers.

Protocol Summary

• Purpose: To enable secure communication between AI agents and merchants during every step of a transaction, thereby establishing a foundational framework for agentic commerce.
• Key Function: It helps merchants verify legitimate agents and filter out malicious bots, addressing a growing problem as AI-driven traffic to U.S. retail websites has surged over 4,700% in the past year.
• Mechanism: The protocol allows approved AI agents to securely transmit crucial details to merchants using cryptographic signatures. This information can include the agent’s intent to buy, data showing if the consumer has an account with the retailer, and payment information.
• Technical Foundation: It is built on the HTTP Message Signature standard.
• Industry Approach: While the initial specifications apply to the Visa network, the company is pursuing an open, ecosystem-wide approach and is aligning with standards bodies like IETF, OpenID Foundation, and EMVCo.
• Industry Support: Visa has worked with numerous partners, including Adyen, Coinbase, Microsoft, Shopify, and Stripe, who provided feedback during the protocol’s development.

The Trusted Agent Protocol addresses a core challenge: the rise of automated transactions has led to issues for merchants, such as bot detection systems blocking valid agent activity and a loss of visibility into the human consumer behind the AI. According to Visa’s Chief Product & Strategy Officer, Jack Forestell, the new protocol is focused on creating “no-code functionality for merchants to securely identify agents with an intent to buy”.

Beyond immediate security needs, the rise of agentic commerce—shopping powered by AI agents acting on our behalf—represents a seismic shift in the marketplace, fundamentally rethinking the shopping experience. This is not just an evolution of e-commerce; it is a move toward a world where AI anticipates consumer needs, navigates options, negotiates deals, and executes transactions autonomously, aligning with human intent.

The stakes are high. By 2030, McKinsey research projects the US B2C retail market alone could see up to US$1 trillion in orchestrated revenue from agentic commerce, with global projections reaching as high as US$3 trillion to US$5 trillion. This shift has the potential to move faster than prior web and mobile-commerce revolutions.

Agentic commerce transforms a fragmented, stressful journey—like a cross-country move requiring dozens of tools and websites—into something personalized, highly efficient, and coherent. The AI agent acts as a personal strategist, designer, negotiator, and logistics manager, making millions of micro-decisions daily on behalf of the user.

The path to purchase is evolving through three key interaction models:

1. Agent to Site: Agents interact directly with merchant platforms (e.g., a travel agent scanning hotel websites and booking a room).
2. Agent to Agent: Agents transact autonomously with other agents (e.g., a personal-shopping agent negotiating a bundle discount with a retailer’s in-house AI commerce agent).
3. Brokered Agent to Site: Intermediary systems facilitate multi-agent and multi-platform interactions (e.g., a restaurant-booking agent contacting a broker like OpenTable to find a table and apply loyalty discounts).

To support this new era, developers are leveraging advancements in AI, driven by six key tools and developments, including:

• Model Context Protocol (MCP): An interoperability standard that allows AI agents to share context, intent, and data across different models and tools, enabling persistent and structured communication.
• Agent-to-Agent Protocol (A2A): A communication protocol for autonomous agents to coordinate, negotiate, and complete tasks directly with one another, supporting real-time, cross-platform marketplaces.
• Agent Payments Protocol (AP2): Google’s open, payment-agnostic protocol designed to enable agents to make verifiable purchases on behalf of users, utilizing cryptographically signed mandates to ensure an auditable, non-repudiable process.

The fundamental shift from a “human-in-the-loop” model to agent-initiated transactions presents significant challenges for the payments infrastructure. New protocols are needed for delegated authorization and programmable spend policies. The risk stack must evolve from relying on behavioral heuristics to establishing protocol-level trust—verifying the agents themselves, not just the users. This includes developing new fraud detection models and adapting standards to require “Know Your Agent” (KYA) identification.

For businesses, the transition requires fundamental changes, including:

• Innovation in areas like customer engagement and loyalty, by developing agents that proactively suggest products and offers based on inferred intent.
• Renovation of core commerce platforms and in-store point-of-service systems to enable agents to execute structured transactions and synchronize digital and physical customer journey.

In this agent-driven economy, businesses must ask: “Are we building experiences for people or their agents?”. The companies that embrace this moment with vision, investing in flexible architectures and rethinking their models, are poised to shape the future of commerce.