
Coinbase's new MPC wallet infrastructure lets AI agents hold crypto and execute trades autonomously. The first agent, Luna, paid for API credits — a step toward machine-driven commerce onchain.
Coinbase rolled out infrastructure that lets AI agents hold crypto wallets and execute trades autonomously, the company said Tuesday.
The system uses multiparty computation (MPC) wallets stored in a secure enclave. The agent receives a delegated signing key that activates only after the user grants permission through a cryptographically signed on-chain message. The wallet is non-custodial for the agent's keys, though Coinbase holds the user's master key. The agent can trade on Coinbase's exchange and on Base, its Ethereum layer-2 network.
The first agent to use the setup was Luna, a bot that autonomously paid for API credits via the platform. Coinbase framed the move as a shift toward machine-driven commerce – AI paying for compute, data, or subscriptions without a human approving each transaction. The company said the model also opens the door for agents to earn crypto, then spend it on external services.
Security rests on the premise that the agent can only spend from the wallet the user designated, and only within the limits coded into the permission message. The secure enclave isolates the signing key from the agent's own infrastructure, which Coinbase argued reduces the risk of unauthorized drain.
The rollout targets a growing cohort of developers who want AI to operate financial primitives directly. Coinbase said it expects the first wave of uses to involve autonomous market-making, subscription payments for AI tools, and agent-to-agent settlements. The service is available now to Coinbase developers through its developer platform.
The move follows a broader push by exchanges to attract machine-driven volume. Unlike traditional API trading, which still requires a human to manage keys and permissions, the MPC-based design lets agents hold a long-term identity onchain without exposing master keys each time.
Prepared with AlphaScala research tooling and grounded in primary market data: live prices, fundamentals, SEC filings, hedge-fund holdings, and insider activity. Each story is checked against AlphaScala publishing rules before release. Educational coverage, not personalized advice.