Cobo Introduces AI-Focused Agentic Wallet To Enable Secure Autonomous Blockchain Transactions

In Brief

Cobo unveils its AI-powered Agentic Wallet, enabling secure autonomous on-chain transactions with MPC technology, Pact-based controls, and task-specific execution frameworks for reliability.

Digital asset infrastructure provider Cobo has announced the launch of the Cobo Agentic Wallet, described as a next-generation wallet specifically developed for artificial intelligence agents. The system is designed to allow AI-driven entities to independently carry out on-chain transactions while remaining subject to controls defined by users

As AI agents increasingly interact with financial assets, attention has shifted from their technical ability to execute transactions toward ensuring that autonomy does not undermine oversight and reliability. The new wallet framework aims to address this balance between operational independence and user governance.

The solution introduces a set of innovations intended to redefine how AI agents interact with blockchain environments. Among these is a non-custodial wallet architecture based on Multi-Party Computation. Traditional approaches to agent-enabled wallets often depend on Trusted Execution Environments, API credentials, or externally owned accounts, which rely heavily on software-based protections.

These methods can expose assets to risk if systems are compromised, credentials are leaked, or language models are manipulated. By contrast, the new system distributes and encrypts private key components, requiring joint participation from multiple parties before a transaction can be authorized. This structure is designed to prevent any single compromised element from independently initiating transfers, offering a mathematically enforced security mechanism rather than relying solely on procedural safeguards.

Framework For Controlled Autonomy And Secure Execution

Another key component is a protocol referred to as Pact, which governs the interaction between users and AI agents. Each operation initiated by an agent begins with a defined agreement outlining permitted actions, operational boundaries, and conditions for completion. These agreements are generated dynamically and enforced at the infrastructure level, expiring once tasks are completed. This approach aims to allow agents to function autonomously within clearly specified limits, replacing static permission models with programmable, task-specific controls.

The platform also incorporates a modular system known as Recipes, which provides predefined execution frameworks for various blockchain activities. These templates are intended to guide agents through complex operations such as lending, token swaps, recurring investment strategies, and payment processing. By relying on verified pathways, the system seeks to reduce errors commonly associated with AI-driven execution, including incorrect contract interactions or inaccurate parameter selection.

Together, these features are positioned as enabling a more reliable method of delegating blockchain tasks to AI systems. The wallet is designed for integration with widely used AI development frameworks and supports a broad range of blockchain networks and digital assets. Operationally, it offers different modes tailored to varying risk levels, including one requiring user co-authorization for transactions and another intended for high-frequency, lower-risk activities.

The development reflects a broader industry trend toward enabling AI agents to operate with financial autonomy while maintaining mechanisms for accountability and control. According to the company, this balance is critical as AI systems evolve from analytical tools into active participants in digital economies.