- World extends AgentKit for verified AI agents using World IDs.
- AI agents can act online on behalf of verified human users.
- The system aims to prevent bots while enabling reliable automation.
World is expanding access to AgentKit, a framework designed to create human-verified AI agents and allow individuals to associate those agents with verified World IDs.
This system allows AI agents to act on behalf of users on the Internet while maintaining identity verification across the World Network.
These developments come as AI agents become increasingly capable of performing online tasks, including shopping, making reservations, navigating websites, and interacting with digital services.
These increasing capabilities have made it difficult for enterprises to distinguish between agents representing real users and networks of automated bots.
AgentKit is positioned as a response to that problem by connecting AI agents directly to World ID, allowing websites and applications to see when agents are working on behalf of unique humans.
The framework is designed to support task delegation while maintaining safeguards around identity verification and user control.
How AgentKit connects AI agents to verified identities
To start using AgentKit, individuals will need a verified World ID, access to the World App, and a supported AI agent, including tools such as Claude Code, Codex, Cursor, Hermes, or OpenClaw.
Users connect human attestations through World’s ToolRouter interface, generate API keys, and connect AI agents in minutes.
Once connected, the agent can interact with services that support AgentKit and perform actions on the user’s behalf.
The system is designed to allow individuals to delegate digital tasks to AI agents while retaining control linked to their verified identity.
According to the framework description, this structure is intended to ensure that AI activity remains attributed to real users rather than anonymous or automated systems.
The demo shows a real-world use case.
This technology was recently demonstrated with the launch of a limited edition of 500 “Human in the Loop” hats, available only to verified World ID holders.
During the demonstration, an AI agent discovered drops, verified eligibility, navigated the store, and completed purchases on behalf of users, all while maintaining a per-person item limit associated with a verified identity.
All 500 hats have been claimed by verified individuals from several countries, including the United States, Germany, Japan, and the United Kingdom.
This demo was used to demonstrate how an AI agent can execute real transactions while maintaining identity-based constraints designed to limit abuse.
This example highlighted how businesses can allow AI agents to complete tasks on behalf of users while preventing abuse by bot networks.
Building a trust layer for the agent economy
As more services integrate AgentKit, World aims to create a layer of trust for the emerging agent economy.
The goal is to enable AI agents to transact and interact online while maintaining accountability to the humans they represent.
The system is intended to support an increasingly diverse set of use cases where AI agents operate autonomously, but within the framework of verified identity and user authentication.
This includes both commercial applications and broader digital service interactions.
The World project was originally conceived by Sam Altman, Max Novendstern, and Alex Blania and aims to provide evidence of human, financial, and connectedness for all humans in the AI era.
The company says AgentKit is part of a broader effort to support identity verification in an environment where AI agents are increasingly able to act independently on online platforms.