World, the identity-focused startup backed by Tools for Humanity, has introduced a new verification tool called AgentKit, aimed at ensuring that real humans are behind AI-powered shopping agents. The launch comes as “agentic commerce”—where AI programs browse and make purchases on behalf of users—rapidly gains traction across the internet.
AgentKit is designed as a software development kit that enables websites to integrate a “proof of human” system into their platforms. This allows businesses to confirm that an AI agent making a purchase is acting on behalf of a verified individual, rather than being part of large-scale bot activity or fraudulent automation.
The system is built on World ID, the company’s core identity framework. In its most secure form, World ID is generated through an iris scan using the company’s Orb device, which converts biometric data into a unique, encrypted digital identifier. This ID can then be linked to AI agents, enabling them to carry verified human authorization while interacting with online platforms.
AgentKit also integrates with the x402 protocol, a blockchain-based standard developed with Coinbase and Cloudflare, allowing AI agents to conduct transactions autonomously while still signaling that their actions are approved by a real user. This combination of identity verification and automated payments is intended to create a more secure and trustworthy ecosystem for AI-driven commerce.
The launch addresses growing concerns around fraud, spam, and misuse as AI agents become more capable of executing tasks at scale. Without proper safeguards, a single bad actor could deploy thousands of bots to manipulate systems, create fake demand, or exploit e-commerce platforms. By tying each agent’s activity back to a verified human, World aims to introduce accountability and reduce such risks.
As major companies continue to invest in AI-powered shopping and automated transactions, the need for trust and verification mechanisms is becoming increasingly critical. World’s new tool represents an early attempt to establish standards for identity and security in this emerging landscape, where human intent and machine execution are becoming closely intertwined.
