Key Takeaways
- World’s newly launched AgentKit developer toolkit incorporates Coinbase’s x402 protocol, developed in partnership with Cloudflare
- AI agents can now transport cryptographic credentials confirming they operate on behalf of verified humans
- Erik Reppel from Coinbase emphasized transforming agents into “legitimate economic participants” instead of questionable automated programs
- Brian Armstrong, Coinbase’s founder, anticipates AI agents will outnumber humans in conducting online transactions in the near future
- Coinbase unveiled an AI agent wallet on its Base blockchain in February for managing autonomous payment flows
The x402 protocol from Coinbase addresses a critical challenge in artificial intelligence: establishing verifiable human identity behind autonomous agents.
World, the identity verification project supported by Sam Altman, introduced AgentKit this Tuesday. This developer toolkit builds upon x402, an open-source protocol that Coinbase and Cloudflare jointly developed. The beta release enables AI agents to carry cryptographic credentials proving human authorization.
The x402 protocol functions by integrating stablecoin micropayments into the fundamental communication protocols of the internet, enabling AI agents and software to execute transactions autonomously.
Erik Reppel, who leads engineering at Coinbase Developer Platform and founded x402, explained the relationship: “Payments are the ‘how’ of agentic commerce, but identity is the ‘who.'”
Coinbase plays an active role in this ecosystem. According to Reppel, platforms implementing the system gain the ability to refuse payments lacking human verification credentials. “As the seller, you can just say, ‘This doesn’t have proof of human attached to it, so I’m going to reject the payment.'”
Coinbase’s Strategic Position in Agent-Driven Commerce
Brian Armstrong, who founded Coinbase, has stated his expectation that AI agents will surpass humans in executing online transactions “very soon.”
Changpeng Zhao from Binance projected an even more dramatic shift, forecasting agents will process one million times the payment volume of humans, “and they will use crypto.”
These projections explain why Coinbase has accelerated its development efforts in agentic payment infrastructure.
The company introduced a specialized wallet for AI agents on its Base blockchain network in February, engineered to process payments while securing private keys within trusted execution environments.
The x402 protocol adds a crucial verification dimension. Beyond providing agents with payment capabilities, it equips platforms with identity confirmation mechanisms.
Addressing the Bot Verification Challenge
Current platform architectures struggle to distinguish between authentic AI agents operating under human authorization and malicious bot networks exploiting system vulnerabilities.
DC Builder, a research engineer at the World Foundation, illustrated the risk: “Think of Ticketmaster: if you delegate an agent the ability to book tickets, you can spawn 100,000 tickets.”
AgentKit solves this challenge by connecting multiple AI agents to a single verified human identity through zero-knowledge cryptographic proofs. Platforms gain the ability to enforce restrictions at the identity level—limiting free trials, capping daily transactions—independent of agent quantity.
A federal judge recently issued a court order preventing AI company Perplexity’s Comet browser from executing purchases on Amazon for users, demonstrating that legal frameworks around agent commerce are actively developing.
Reppel articulated the challenge Coinbase aims to address: “What we need are robust, open ways of understanding which is which — being able to tell when you’re talking to an AI, a human, or a specific human’s AI.”
World’s AgentKit presently employs iris-scanning Orb hardware for biometric authentication, with expansion plans including NFC-enabled passports and government-issued identification documents. The verification network encompasses nearly 18 million confirmed individuals spanning 160+ countries.
