
Singapore-based crypto custody and wallet provider Cobo has officially launched Cobo Agentic Wallet, a new product designed to let AI agents safely perform onchain tasks with built-in human oversight and strong security controls.
The wallet introduces a human-agent authorization protocol and a modular skill framework to ensure reliable and secure execution. Its standout innovation is the “Pact” mechanism — a dynamically generated contract for each task that clearly defines:
These boundaries are enforced at the infrastructure level, not just through policy.
Changhao Jiang, co-founder and CTO of Cobo, explained the philosophy behind the design:
“Most agentic wallets simply hand over wallet capabilities to agents and hope for the best. We give agents a Pact. A Pact defines what an agent can do, where it must stop, and when execution ends — and it is enforced at the infrastructure level.”
To further enhance safety, the wallet uses multi-party computation (MPC). This means that even if an AI agent is compromised, hallucinates, or credentials are leaked, it cannot generate a valid signature on its own. This provides a cryptographically enforced guarantee rather than relying solely on best practices.
Cobo Agentic Wallet supports more than 80 blockchains, including major networks such as:
It also integrates with popular AI development frameworks, including LangChain, OpenAI Agents SDK, Claude MCP, Agno, and CrewAI.
Cobo is not alone in this space. Other major players have been moving quickly into AI-powered wallets:
As AI agents become more capable of managing real capital onchain, the need for secure, controllable execution frameworks is growing rapidly. Cobo’s approach — combining clear task boundaries (Pacts) with MPC security — aims to bridge the gap between powerful AI autonomy and the safety requirements of users and institutions.
The product is positioned as a practical step toward the next generation of onchain AI agents that can “act with real capital, and within boundaries humans can trust.”
