The hedge-fund toolkit is coming to the brokerage app. At its HOOD Summit in Houston this week, Robinhood unveiled Robinhood Agents, an AI experience built directly into its app that can research markets, analyze portfolios, develop trading strategies, and place orders — all inside limits the customer sets. CEO Vlad Tenev framed the launch as the inevitable answer to behavior already visible on the platform: in remarks shared on X, he said traders are increasingly integrating AI into their strategies, and the company "wanted to make that easier."

The launch matters because of what it normalizes. This is one of the first times a mainstream brokerage has let retail customers delegate end-to-end trading decisions to configurable AI agents inside the core app rather than through external bots — moving the industry's biggest unanswered question, who is liable when an AI agent executes a bad trade, from theory into a settings toggle.

What Robinhood Agents actually does#

Setup takes three steps: customers name the agent, open a dedicated agentic trading account, and connect an AI model. The model menu opens with OpenAI and Anthropic, and Robinhood says access to OpenAI's GPT-Luna model will be free through the end of 2026. Once configured, the agent can research markets, analyze a portfolio, develop strategies, and place orders across equities, options, and crypto.

Two companion features widen the surface. Agent Apps pipes institutional-grade data from eleven third-party providers into the agent, so it trades on structured intelligence rather than vibes alone. And Loops, still in development, turns a strategy into standing instructions that the agent re-executes continuously — including overnight.

Robinhood CEO Vlad Tenev speaking onstage at TechCrunch Disrupt
Vlad Tenev, Robinhood CEO, speaking at TechCrunch Disrupt NY 2016. Photo: TechCrunch / Wikimedia Commons (CC BY 2.0).

The guardrails — and the off-switch#

Robinhood built the product with a visible anxiety about what it is unleashing. Agents operate only inside the dedicated agentic account; the company's materials stress they are blocked from reaching the main account balance, so a malfunctioning agent cannot drain a customer's entire portfolio. Trade approvals are switched on by default, requiring a human sign-off for every transaction, and agents cannot borrow on margin at launch.

The catch: the approvals can be switched off entirely. Flip the setting and the agent can trade in a loop without asking — and Robinhood warns that once Loops is enabled, an agent may place, change, or cancel trades on its own. Tenev told CNBC's Squawk Box that algorithms already drive much of the volume on exchanges and that agents open up "a new capability frontier." The frontier, in other words, ships with a toggle.

A small AI robot inside a caution-striped glass box next to a safe, with a human hand near an approval button
Editorial illustration: the dedicated-account guardrail, and the approval switch that can disable it. (AI-generated)

Why now: 150,000 customers were already doing this#

Robinhood Agents did not arrive from nowhere. In May, the company opened its platform to external AI agents through Model Context Protocol servers, letting technically adept customers wire their own bots into Robinhood. Since then, more than 150,000 customers have opened agentic trading accounts, and the company says those agents now use its trading tools nearly 30 million times a day.

That adoption curve explains the framing. The brokerage is not inventing agentic trading; it is absorbing it — moving from watching sophisticated users bolt bots onto the platform to offering a sanctioned, model-agnostic version anyone can configure in minutes. Rival firms that spent the year watching from the sidelines now face the same three questions: whether to offer similar capabilities, what guardrails to demand, and how to price the liability when an agent executes a strategy its owner barely read.

What to watch#

  • The liability question. Manual approvals are the default today, but the industry has no settled answer for who pays when an autonomous loop misfires. Expect regulators to take a close interest.
  • The margin of safety in practice. Dedicated accounts cap the downside, but an agent with approvals disabled can still chew through that account fast. The first ugly incident will define the product's reputation.
  • The always-on endgame. Robinhood also announced 24/7 weekend stock trading — pending regulatory approval, aimed at 2027 — and crypto perpetual futures with up to 10x leverage. An agent that never sleeps paired with markets that never close is where this is headed.

Robinhood's own summary line — "the power of hedge funds in every trader's pocket" — is marketing, but it captures the direction of travel. Algorithmic tooling once reserved for professionals is being productized for retail, one approval toggle at a time. The agents have entered the app; whether they stay invited depends on what happens the first time one of them makes a really expensive decision.