Cobo AI Enhances Agent Control with Structured Permission System

On April 21, Cobo AI's Growth Lead, Brad Bao, discussed advancements in agent control at an event titled 'Decoding Web 4.0: When AI Agents Take Over On-Chain Permissions.' According to BlockBeats, Cobo is restructuring the logic of agent fund control through a structured permission system. In Cobo's products, the relationship between users and agents is defined by an executable 'Policy/Agreement,' rather than a simple wallet authorization. This agreement includes four core elements: Intent, Path, Rules, and Termination. This approach transforms the traditional 'all-or-nothing' permission control into a negotiable, auditable, and revocable fine-grained governance mechanism. Cobo employs Multi-Party Computation (MPC) technology at the execution level to ensure that even if an agent behaves abnormally, the system is attacked, or prompt injection risks occur, users maintain control over their funds. No single entity can bypass constraints to independently transfer assets. Brad Bao emphasized that within this framework, agents have sufficient freedom to complete complex tasks, but their actions are always confined within verifiable boundaries. He stated that this system essentially builds a layer of 'trust and insurance infrastructure' in the agent economy. By combining 'contractual permissions and secure computation,' it allows for controlled delegation of agent actions, enhancing automation efficiency while ensuring user asset security and outcome certainty.