Robinhood Puts AI Trading Agents in 29 Million Pockets

Robinhood is about to let millions of people hand their stock picks to an AI. Starting this week, the brokerage is rolling out AI trading agents powered by models from OpenAI and Anthropic to all of its roughly 29 million customers, letting everyday investors automate research, strategy-building, and trading with plain-English instructions.

The launch of these AI trading agents was timed to the company's annual HOOD summit in Houston, Texas, and turns a feature that began as a power-user experiment into a mainstream product. Back in May, Robinhood released a tool that let technical users connect their own agents to the platform through the open Model Context Protocol standard.

That experiment clearly found an audience. More than 150,000 customers have opened dedicated agentic trading accounts since the May launch, and the company's figures say agents now call Robinhood's tools almost 30 million times a day, according to its summit briefing.

If those numbers hold through the general rollout, Robinhood will become the first broker to put a non-technical AI trading agent in front of tens of millions of retail investors — a step that could reshape how ordinary Americans invest and send ripples through markets more broadly.

How Robinhood's AI trading agents work

Setting up an agent starts with two choices: a name and a model. In a demonstration viewed by Fortune, a user named their agent and then picked between OpenAI's GPT-6 Luna or GPT-6 Sol, or Anthropic's Opus 4.8.

Once configured, the agent takes plain-English instructions. Simple requests like "Buy $200 of Ford stock" sit alongside more ambitious assignments the company calls Loops. A Loop lets an agent check the market every morning and execute a trade when certain conditions are met, or run a continuous overnight strategy to look for opportunities while the user sleeps.

That flexibility is the real product here. Rather than a static robo-advisor questionnaire, the agent combines market analysis, watchlist building, and trade execution in one conversational loop across stocks, options, and crypto.

Model choice matters because pricing differs. According to coverage of the launch, usage of OpenAI's GPT-6 Luna will be free until the end of the year, while other models carry their own costs. Agents, the feature the company calls Robinhood Agents, is coming soon to eligible customers in the United States.

Guardrails built for real money

Handing an autonomous system access to a brokerage account is the kind of thing that keeps compliance teams up at night, and Robinhood has clearly thought about that. The company described a layered set of guardrails designed to keep agents from behaving unexpectedly.

The centerpiece is a dedicated agentic trading account. The agent can only use the money inside that account, so a runaway strategy cannot drain a user's main portfolio.

Users can also set limits on how much the agent can trade at a time. And manual approval of each trade is on by default, meaning the agent proposes and the human disposes — though customers can change that setting if they want more autonomy.

That approval-first design echoes a broader industry pattern: agent platforms are converging on human-in-the-loop defaults while leaving the door open for experienced users to loosen the reins. It is a sensible posture for a product that will soon be in the hands of first-time investors.

Why this is bigger than one brokerage

Robinhood did not stop at agents. The same summit brought crypto perpetual futures with leverage of up to 10 times, extended options trading hours, increased intraday margin, and weekend trading in select US stocks and ETFs through the Bruce ATS platform — an aggressive push to capture more activity from frequent traders.

But the trading agents are the headline, because they signal a shift in who agents serve. Until now, agentic finance has mostly meant APIs, developer SDKs, and institutional pilots. Robinhood is betting that the mass-market moment is here — that millions of people are ready to describe a strategy in words and let a model execute it.

The bet fits a pattern playing out across the agent economy. Meta recently launched Muse for small business with connectors into everyday work apps, and OpenAI's always-on Dots agents are pushing into enterprise workflows. Finance, with its real-money stakes, is the most consequential frontier yet.

The open question is behavior, not technology. What happens when millions of agents wake up each morning, check the same markets, and react to the same signals? Individual guardrails protect individual accounts, but correlated agent behavior across millions of users is a market-structure question nobody has stress-tested.

Regulators are watching the clock too. The push toward longer and weekend trading hours has already drawn attention from market officials who warn that acting outside traditional hours can mean missed opportunities or extra risk.

For now, the experiment is live and the numbers are large. With younger investors already trusting AI at surprising levels, Robinhood's 29 million customers may be exactly the audience that says yes to a trading agent first — and asks questions later.

Sources: WDC TV News, reporting on the Fortune demonstration; TradersUnion, citing CoinDesk's summit coverage; AI StockWire's breakdown of the HOOD summit announcements.