Tools for Humanity (TFH), the startup behind the World identity platform co-founded by Sam Altman, has released a beta version of AgentKit – a software development kit aimed at commercial websites seeking to verify that a real human authorises the purchasing decisions made by AI agents acting on their behalf.
The release comes as agentic commerce, the use of AI programmes to browse websites and execute purchases autonomously, accelerates across major platforms. With that growth has come a corresponding rise in concerns about fraud, spam, and large-scale automated abuse.
How AgentKit works
AgentKit integrates World ID, TFH's biometric identity credential, with the x402 protocol, a blockchain-based open standard developed by Coinbase and Cloudflare that enables automated programmes to transact directly online without requiring human intervention at each step. Users register their AI agents with their World ID, which then signals to participating websites, via the x402 system, that a verified and distinct human has authorised the agent's activity.
The most secure form of World ID is generated through a biometric scan of a user's iris via TFH's proprietary Orb device. The Orb converts the iris scan into an encrypted digital code, which can be accessed through the World app. AgentKit requires this Orb-derived credential for the highest tier of verification.
TFH described AgentKit as a complementary extension to the x402 v2 protocol, developed in coordination with Coinbase. The company noted that any website already operating on x402 can enable proof-of-human verification alongside, or instead of, micropayments. TFH's chief product officer compared the mechanism to delegating power of attorney: the World ID badge confirms that a real, unique individual stands behind the agent's decisions, while websites retain the ability to block users they consider to be acting in bad faith.
Industry context
The launch aligns with a broader industry shift toward agentic infrastructure. In 2024, Amazon and Mastercard each introduced automated buying capabilities to their respective platforms, and Google subsequently launched its own protocol to support agentic transactions. As the ecosystem matures, the question of human accountability in AI-driven commerce has moved to the foreground of industry debate.
TFH's move positions World ID as potential infrastructure for that accountability layer. By anchoring agent behaviour to a verified human identity, the system attempts to address a structural gap: existing payment and commerce frameworks were not designed with autonomous AI actors in mind, and the x402 protocol itself lacks a native mechanism to confirm human intent.
AgentKit is currently available to developers in beta, with TFH indicating that feedback will shape future iterations of the tool. Whether adoption follows will depend in part on how quickly the x402 protocol gains traction among commercial platforms, and how the industry weighs biometric verification against privacy considerations at scale.
