World Expands Agentkit Access as AI Agents Execute Purchases Across Four Countries
World, the Sam Altman-backed project, announced on June 24 that it is expanding access to Agentkit — a development framework that allows individuals to connect AI agents to a verified World ID. The move aims to address the growing challenge of distinguishing legitimate AI agents from malicious bot networks.
Agentkit was initially unveiled in March through a beta version developed in collaboration with Coinbase and Cloudflare, integrating World’s biometric identity layer with the x402 protocol for secure stablecoin micropayments. In April, the infrastructure was embedded into the World ID 4.0 upgrade, establishing integrations with platforms like Vercel, Okta, and Box.
According to a statement, the latest expansion provides an identity-based trust layer, enabling businesses to verify that each agent corresponds to a unique, verified individual. Users can set up the system with a verified World ID, the World app, and a compatible AI agent platform such as Claude Code, Codex, Cursor, Hermes, or Openclaw. The process involves linking a digital proof of humanness through World’s Toolrouter interface to generate an API key, which the company says takes only a few minutes.
World conducted a pilot program featuring a limited release of 500 ‘Human in the Loop’ hats restricted to verified World ID holders. During the trial, AI agents independently tracked the product launch, verified eligibility, navigated the storefront, and completed transactions. The integration enforced a one-item-per-person limit across orders from the US, Germany, Japan, and the UK. Company officials noted that the trial demonstrates how e-commerce platforms can safely allow automated buyers without risking inventory depletion by resale bots.
As Agentkit integration expands to broader digital services, World aims to build an accountable ‘agent economy,’ ensuring autonomous software remains bound to and controlled by the humans it represents.