MetaMask, the popular self-custodial wallet owned by Consensys, has officially unveiled its Agent Wallet, a new product designed specifically to allow AI agents to autonomously interact with decentralized finance (DeFi) protocols while maintaining strict user-defined security barriers. The launch, announced on Monday, places MetaMask at the forefront of a growing movement to equip AI-driven software with the ability to trade, provision liquidity, and manage digital assets on behalf of human users—without surrendering control of private keys.
The Agent Wallet supports a wide range of activities across Ethereum Virtual Machine (EVM)-compatible blockchains and Hyperliquid, including token swaps, perpetual futures, prediction markets, and liquidity provisioning. MetaMask is pitching its security-first architecture as the key differentiator in a field where many competing projects grant AI agents raw access to private keys, a practice that Consensys executives warn could normalize the very custodial risks that decentralized finance was built to eliminate.
Currently, the product is available to roughly 200 early-access users, with a broader public rollout planned for this summer. According to MetaMask Senior Director of Product Zhen Yu Tong, speed of infrastructure development is critical. “It’s genuinely day one for agents, but the infrastructure decision can’t wait because agents are already touching real money, and most of them are doing it the wrong way,” he said in an interview.
Every transaction initiated by an agent goes through multiple layers of automated defense before execution: transaction simulation, threat scanning powered by Blockaid, clear signing, and MEV protection via Servo. If a transaction is flagged as malicious, it cannot proceed without explicit human approval triggered by two-factor authentication. To further reassure users, MetaMask extends its Transaction Protection program—which covers losses up to $10,000—to any transaction that passes its safety checks.
Users can choose between two operating modes. The default “Guard Mode” enforces spending limits, protocol allowlists, and other restrictions, requiring 2FA for anything outside these bounds. An opt-in “Beast Mode” reduces the frequency of prompts, allowing agents to act more freely within predefined guardrails, but still mandates human approval for any transaction flagged as potentially harmful. As Tong explained, “Beast Mode is for users who want genuinely hands-off operation… What Beast Mode does not do is switch off the safety net. If our threat detection flags a transaction as malicious, 2FA still fires no matter what mode you’re in. That’s non-negotiable.”
The wallet’s design acknowledges a fundamental vulnerability of large language models (LLMs): prompt injection attacks, where malicious instructions can trick an AI into unintended actions. By never exposing private keys directly to the agent and keeping them inside a hardware-isolated trusted execution environment supplied by Cubist, MetaMask aims to contain the fallout even if the underlying AI is compromised. Consensys CEO and Ethereum co-founder Joe Lubin framed the launch in sweeping terms: “The next great expansion of the onchain economy won’t be driven by humans alone. Agents will manage real capital and make real financial decisions, and the infrastructure underneath has to be worthy of that.”
The Agent Wallet arrives amid a flurry of activity around AI agent infrastructure in crypto. In February, Coinbase introduced Agentic Wallets with a similar focus on secure private key isolation. MoonPay followed in March with Ledger hardware wallet integration for human-approved AI transactions, and later co-launched the Open Wallet Standard alongside contributors including PayPal, the Ethereum Foundation, Solana Foundation, Ripple, and Base. MetaMask’s entry, with its emphasis on transaction simulation and mandatory security chokepoints, adds a major player to this evolving landscape.
