MetaMask has launched a self-custodial wallet designed for AI agents. The company disclosed that this product enables automated software to execute transactions and manage funds in decentralized finance scenarios, while users retain full control over their assets.

Supports multiple types of on-chain transactions
The product, named MetaMask Agent Wallet, enables AI agents to perform various operations on Ethereum-compatible blockchains, including token swaps, perpetual contracts, prediction markets, and liquidity provision.
The product is currently available through a limited early access program. A broader rollout is expected to proceed over the coming months.
Security design is a core selling point.
The company positions security as the primary feature of this wallet. As disclosed, every transaction initiated by the AI agent undergoes transaction simulation, threat scanning, and MEV protection before execution.
If the system detects suspicious or malicious risks, the transaction will be escalated to manual review, and the user will be required to complete two-factor authentication to proceed.
MetaMask also states that transactions deemed secure may be covered under its Transaction Protection program, with coverage of up to $10,000 for potential losses.
Users can set the control intensity.
In terms of usage, users can select the default "Guard Mode," which enables spending limits, a protocol whitelist, and approval requirements to restrict the AI agent's operational scope.
An alternative mode is "Beast Mode." This mode reduces interactive prompts but still requires manual approval for transactions that may pose malicious risks.
As AI agents become more involved in the crypto market, related tools are shifting from information assistance to actual trade execution and fund allocation. MetaMask’s launch of this new wallet reflects wallet providers’ efforts to build a more comprehensive infrastructure for AI-driven automated trading.

Additional information: Joe Lubin, CEO of Consensys and co-founder of Ethereum, stated that the expansion of the on-chain economy in the future will not be driven solely by human actions; AI agents will directly manage real funds and make financial decisions.

