AI agents comparing prices, booking tables, and making payments on your behalf — McKinsey projects this market could reach $3 to $5 trillion by 2030. But there's a problem: nobody can verify whether there's a real person behind these agents or not.

TL;DR
AI agent shopping boom Can't tell bots from humans World ID iris verification Delegate to agents via AgentKit Sites verify "real human" behind agent

What Is This?

World (formerly Worldcoin), the identity project co-founded by Sam Altman, just launched AgentKit. In short, it's a developer toolkit that gives AI agents cryptographic proof there's a real person behind them.

Here's how it works: you scan your iris with the Orb device to create a unique World ID. Link that ID to your AI agent, and when the agent accesses a website, it can prove "I'm an agent sent by a verified, unique human". It uses zero-knowledge proofs, so no personal information is ever exposed.

AgentKit integrates with the x402 protocol developed by Coinbase and Cloudflare — a layer that enables AI agents to make micropayments. AgentKit adds "proof of human" on top. As Erik Reppel, head of engineering at Coinbase Developer Platform, put it: "Payments are the 'how' of agentic commerce, but identity is the 'who'".

Key Concept

KYC (Know Your Customer) → KYA (Know Your Agent) → Know Your Human. Identity verification for agentic commerce is evolving through three stages.

The World network currently has approximately 18 million verified users across 160+ countries. Bain estimates AI agents could account for up to 25% of all U.S. e-commerce by 2030.

What Changes?

Until now, websites just blocked all automated traffic — legitimate agents and malicious bots alike. AgentKit changes the rules of the game.

Current ApproachWith AgentKit
Bot handlingBlock all automated trafficAllow human-backed agents only
Fraud preventionMicropayments as rate limiterLimit agents per unique human
ReservationsCAPTCHA, phone verification1 unique human = 1 booking guaranteed
PrivacyPayment trail tracks everythingZero-knowledge proofs preserve anonymity
Free trialsUnlimited email account abuseN trials per unique human

For example, if a popular restaurant platform adopts AgentKit, AI agents can still make reservations — but one person running 100 agents to hoard seats becomes impossible. Same goes for concert ticket scalpers. The bottom line: this isn't about blocking agents, it's infrastructure for distinguishing good agents from bad ones.

According to PYMNTS research, 56.3% of global companies face bot- or agent-related threats, losing an average of 3.1% of annual revenue to digital identity system gaps — $94.9 billion in aggregate.

Getting Started

  1. Create a World ID
    Find your nearest Orb location at world.org and complete the iris scan.
  2. Check the AgentKit docs
    Access the developer preview at docs.world.org/agents/agent-kit. Currently in beta.
  3. Delegate World ID to your agent
    After verification, register your agent to carry "proof of human" on x402-enabled sites.
  4. Integrate into your platform (for businesses)
    If you already use x402, you can add AgentKit as an additional layer — require micropayments and proof of human together, or choose one.
  5. Set limits
    Configure per-human caps: number of agents, daily transactions, free trial uses, etc.

Note

AgentKit is currently in beta with Orb-based biometric verification required. NFC passport/ID verification is planned but not yet available.