Sam Altman’s World has unveiled a new developer toolkit designed to help websites and online services verify that artificial intelligence agents are operating on behalf of real people rather than autonomous bot networks.
The toolkit, called AgentKit, launched today in beta and integrates World’s identity verification system with x402, an open protocol developed by Coinbase and Cloudflare. The technology allows AI agents to carry cryptographic proof that they represent a unique human while interacting with websites, application programming interfaces, and digital services.
The release targets a growing problem in AI commerce. As agents begin handling tasks such as booking services, comparing prices, and completing purchases, websites need ways to distinguish legitimate automated activity from spam, bot swarms, and abuse.
McKinsey estimates agentic commerce could generate $3 trillion to $5 trillion globally by 2030, while Bain has projected AI agents could account for as much as 25% of US ecommerce sales by the end of the decade.
Under the model, a verified user can delegate their World ID to an AI agent, enabling that agent to prove that a real, unique human backs it without exposing the person’s actual identity.
That gives websites a privacy-preserving signal they can use alongside payment checks to decide whether to grant access, allocate resources, or enforce one human one account style limits.
Coinbase and Cloudflare introduced the x402 Foundation last year to push x402 as an open standard for internet native payments, especially for AI agents paying for content, APIs, and digital services. Coinbase describes x402 as a protocol that enables stablecoin payments directly over standard web requests, effectively turning payment into a native part of internet traffic.
World’s pitch is that payments alone do not prove uniqueness. A single operator could still run thousands of fee-paying agents, which means micropayments can limit spam but not verify how many real people are behind the activity.
By attaching proof of unique human to x402-compatible requests, AgentKit gives developers a second trust signal that could make AI agents more acceptable to websites that currently block most automated traffic by default.
The launch also plugs into a broader effort across tech and commerce platforms to prepare for AI-driven transactions. World says its network now spans nearly 18 million verified humans in more than 160 countries. That gives it a sizable installed base for developers who want identity checks without collecting full personal data.
