What AI Trading Agents Are Up to on Robinhood

In the weeks since Robinhood opened its doors to AI agents, customers of the popular financial app have been using agents for everything from research to trading. Now, a key question for Robinhood’s broader revenue opportunity is how much of an edge the technology can give investors.

In late May, Robinhood started letting customers connect their own AI agents to its app through the Model Context Protocol, a standard that lets authorized agents use outside data and services. That followed the rise of popular AI models like Claude Opus, which Anthropic touted for work such as financial analysis.

Once set up with a dedicated brokerage account, agents can analyze holdings across a user’s Robinhood accounts. They can also carry out a trading strategy with a set amount of funds in those dedicated accounts without a human approving each transaction. Robinhood has said customers opened more than 70,000 accounts for their agents as of early July. That’s rapid growth, but still a small number compared to its overall customer base of 27.7 million. 

So far, Robinhood customers are largely treating agentic accounts as a place for experimentation rather than a replacement for their primary accounts. Some appear to be using the agentic accounts primarily for research rather than for placing trades, according to Abhishek Fatehpuria, vice president of product management for brokerage at Robinhood Markets. 

Still, some interesting trading behavior is emerging. In some cases, people are using their agentic accounts to compare an agent's performance with their own, or they’re pitting different AI agents against each other to see how they do, he said. 

One potential question for Robinhood shareholders, assuming agentic trading takes off, is how valuable those automated trades will be as a source of revenue. Though the company has been branching into subscriptions and other financial services in recent years, revenue from customers’ trading activity made up the majority of its revenue in the first quarter. It doesn’t charge commissions on stocks, ETFs or equity options—it generates revenue from payments it receives when it sends its customers’ orders to trading firms and exchanges.

From the perspective of those trading firms, payments for orders typically reflect how much they expect to earn from executing the orders. The firms aim to profit partly from the spread between buying and selling prices, and that strategy tends to work best when retail customers’ trades are unlikely to signal a broader market move. Trades in highly liquid stocks might command lower payments, for example, because spreads are tighter and there’s less room for market makers to profit. 

Arguably, if AI systems can trade on new information more quickly than humans, their orders could be harder for market makers to profit from. If that’s the case, agentic trading activity might generate lower payments for Robinhood than regular, human-driven retail order flow. On the other hand, some agent-driven orders might well be more attractive if the agents’ strategies are easy to anticipate. Right now it’s likely hard for the firms involved to make any judgments about how agents are doing at trading, given how new the feature is. 

Meanwhile, Robinhood has quickly expanded the trading functions and data agents can use, so agentic strategies are evolving. The feature offered just stock trading to start but soon added options trading. The company said earlier this month it plans to add crypto for agentic accounts soon. Other offerings like event contracts could follow. 

Robinhood has also been giving agents access to more and more types of data, including technical indicators, tax lots and company earnings, that can inform trading strategies. 

At the same time, the agentic trading feature could overlap with some of Robinhood’s other products. For instance, Robinhood offers AI-generated market and portfolio summaries included in its Gold subscription. Robinhood has also been expanding beyond active trading to more types of investing, so growth in the new agentic accounts might reshuffle that mix. 

Robinhood’s retirement accounts held about $27.4 billion at the end of the first quarter. The company added an automated investing service last year that manages portfolios for a small percentage fee, which attracted more than 285,000 customers and $1.6 billion in assets by late April. And last month, Robinhood began rolling out a network connecting customers with independent human financial advisers.

Robinhood reports second-quarter results next Wednesday. Stock and options activity has been trending higher than in the year-ago period, based on the company’s most recent monthly activity disclosures. 

But Robinhood’s growth in trading revenue lagged customer trading activity in the first quarter. Executives on that earnings call said that take rates can decline as more activity comes from frequent traders, adding that the company focuses on attracting active traders and gaining market share rather than maximizing revenue per trade. In filings, Robinhood said that a shift toward tickers that generate lower payments weighed on equities and options revenue growth in the period. 

Similarly, agentic trading could benefit Robinhood if it adds to customers’ overall activity, even if each trade generates less revenue. And Robinhood’s “bring your own agent” approach means that traders or outside AI providers, not the company, bear the cost of running the underlying models. At a time when businesses are scrambling to rein in their own AI expenses, that low-cost way to experiment may be one of agentic trading’s biggest attractions for Robinhood and its shareholders.

New From Our Reporters

Exclusive

Anthropic in Talks to Add Billions to Bank Credit Line Ahead of IPO

By Valida Pau

Sunday Insights

Alphabet, SpaceX Spur Record U.S. Equity Sales

By Valida Pau