Human.tech has introduced Agentic Wallet as a Protocol (WaaP), a natural-language wallet infrastructure designed for AI agents while keeping humans in ultimate control through cryptographic enforcement. The system dissolves the traditional wallet interface into natural language interactions, enabling agents to trade, manage portfolios and execute blockchain operations within strict, cryptographically enforced boundaries.
How It Works: Privileges and Human-in-the-Loop
According to the official announcement, WaaP is powered by "privileges" that set spending caps, time limits and approved addresses. A policy engine adds human-in-the-loop approvals for higher-risk actions, with Telegram prompts allowing one-tap confirmations. "The wallet is no longer something you open. It is something that acts for you," said Shady El Damaty, co-founder of Human.tech. "But that only works if the human remains root authority."
Two-Party Computation Custody Model
The core of the system is a two-party computation (2PC) custody model that splits private keys between a user’s device and a secure enclave. This ensures that neither agents nor developers can act independently. The model extends beyond crypto wallets to broader secret management—including bank accounts, API keys and root privileges—with future integration planned for decentralized key management via Ika Network.
Regulatory Tailwinds and Ecosystem Growth
The launch comes as the agent economy accelerates, driven by frameworks such as Coinbase Agentkit, Langchain and CrewAI. Regulatory pressure is also mounting: the European Union’s AI Act takes effect in August 2026, mandating human oversight in autonomous systems. Human.tech said Agentic WaaP is designed to meet these requirements. The broader ecosystem already includes Human Passport, with over 3 million users and 175 partners, securing more than $500 million against Sybil attacks and backed by $3 billion in restaked ether. Agentic WaaP is now available to developers without API key requirements, with plans to expand integrations across agent frameworks.