Robinhood has already seen demand for a more technical version of the idea. More than 150,000 agentic accounts have been created since it opened its trading infrastructure to outside AI agents earlier this year, according to the company, generating millions of calls to its tools each day.

The embedded version removes much of that technical setup.

It also raises an obvious question: What happens when an AI model gets a trade wrong?

“I would think of the agent as an extension, like a trading tool that’s mainly for research that can execute trades,” Gina Pasqua-Abeles, senior director of product management, said. “We do expect users to still have some level of oversight over what their agent is doing.”

Robinhood will default to requiring customers to approve every trade an agent proposes. Users can turn that protection off. Agents cannot borrow on margin at launch and can only access money placed in their dedicated account.

Robinhood is also selling access to outside data that agents can incorporate into their decisions, including options information from Unusual Whales, market data from Nasdaq, crypto data from Token Terminal and government activity tracked by Quiver Quantitative.

The company’s push into crypto-style trading goes further with perpetual futures.