World and Coinbase Launch AgentKit — A New Trust Layer Providing 'Proof of Human' for AI Agents

World and Coinbase Unveil AgentKit to Verify AI Agent Identity

Coinbase and World Launch AgentKit to Verify Humans Behind AI Agents

Coinbase and Sam Altman's World launch AgentKit to verify human intent behind AI agents and enable trusted digital transactions.

On March 17, 2026, World (formerly Worldcoin), co-founded by Sam Altman, and Coinbase announced the beta release of "AgentKit," an identity verification toolkit for AI agents, on their official blog. AgentKit is the industry's first developer toolkit that cryptographically proves a real human exists behind an AI agent.

AgentKit integrates with the x402 protocol, a payment protocol co-developed by Coinbase and Cloudflare, building a trust foundation for the AI agent economy through both "payments" and "identity verification." It is currently available as a limited beta for developers who hold a World ID.

Background and Industry Trends

The era of AI agents acting autonomously online has arrived. Scenarios where AI handles tasks that humans once performed at their keyboards — restaurant reservations, price comparisons, product purchases — are rapidly increasing.

According to McKinsey's projections, the market for agentic commerce (a model where AI agents conduct purchases, negotiations, and transactions on behalf of humans) is expected to reach $3-5 trillion globally by 2030. Bain further estimates that AI agents will handle up to 25% of U.S. e-commerce transactions by the same period.

However, the current web faces a significant problem. Most websites uniformly block all automated traffic because there is no way to distinguish between malicious bots and legitimate AI agents acting on behalf of users. This is precisely the backdrop driving calls for Know Your Agent frameworks to verify agent identity. This situation represents a serious bottleneck for the development of agentic commerce.

How AgentKit Works — A Two-Layer Structure of "Payment" and "Proof of Human"

At the core of AgentKit is the concept of "proof of unique human." While Google's UCP aims to standardize commerce, AgentKit focuses on standardizing "trust." Users who have verified their identity with World ID delegate their World ID to their AI agents. When an agent accesses an x402-enabled website, the site can request "payment," "proof of human," or both. If the agent holds valid proof, access is granted.

Crucially, a single user can delegate their World ID to multiple agents. However, the website can recognize that those agents are tied to the same person. This means that even if someone deploys hundreds of agents to attack a platform, the site can detect it as "actions by a single human" and impose restrictions.

Erik Reppel, Head of Engineering at Coinbase Developer Platform and founder of Coinbase x402, told Coinpaper: "Payments are the 'how' of agentic commerce, but identity is the 'who.' By integrating World ID with the x402 protocol, developers get a complete trust stack."

The x402 protocol itself is growing rapidly. According to a joint announcement with Cloudflare, it has processed over 100 million micropayments in just six months since its 2025 launch. However, payments alone don't solve the identity problem. If economic incentives are high enough, a cost of a few cents per request is negligible for attackers. AgentKit adds a trust layer of "human uniqueness" on top of this payment layer. By combining payment tokenization with identity verification, a more robust trust stack is achieved.

Impact and Actions for E-Commerce Businesses

AgentKit's impact on e-commerce businesses spans multiple dimensions.

Accepting agent-mediated purchases becomes realistic. Of the automated traffic previously blocked uniformly as bots, it becomes possible to selectively accept only "legitimate agents backed by humans." TechCrunch reporting has also identified trust assurance for AI shopping agents as a critical challenge for the retail industry.

Fraud prevention is another key benefit. Issues such as scalper bots hoarding reservations, free trial abuse, and review manipulation can be addressed by imposing "one limit per human." World's official blog describes a case on the agent-driven platform "Moltbook" where a small number of individuals deployed large numbers of agents to distort engagement metrics.

Privacy protection is also a distinguishing feature. World ID leverages zero-knowledge proofs, certifying only the fact that "this person is a real, unique human" without disclosing any personal information. The ability to ensure trust while protecting customer privacy is a significant advantage for e-commerce businesses.

At this stage, AgentKit is a limited developer beta and is not yet ready for immediate adoption by e-commerce businesses. However, World has approximately 18 million verified users across more than 160 countries, and a formal version 1.0 release is planned alongside the rollout of the next-generation protocol (World ID 4.0).

Summary

AgentKit is a concrete solution to the fundamental challenge of "lack of trust" in the AI agent economy. The "complete trust stack" combining Coinbase's x402 for "payments" and World's "proof of human" has the potential to become essential infrastructure for the widespread adoption of agentic commerce.

What e-commerce businesses should watch now is the timing of the official releases of x402 and AgentKit, and how major e-commerce platforms like Shopify and Amazon will incorporate agent authentication. As we move toward an era where AI agents are treated not as "suspicious automated traffic" but as "legitimate economic participants," the time has come to begin considering your own agent-readiness strategy.