Robinhood AI Trading Agents Put Automated Trades in Reach

Robinhood is bringing AI trading agents directly into its app, with access planned for its roughly 29 million customers starting the week of its September 29, 2026 announcement. Users can choose OpenAI or Anthropic models and ask agents to research, develop strategies, or execute trades, moving agentic trading beyond the technical setup required by Robinhood’s earlier MCP tool. The new service and its planned customer reach

Loops turn instructions into recurring activity

The difference from a chatbot that suggests a trade is that Robinhood’s agents can act through trading accounts. A user might tell one to buy $200 of Ford stock, or configure a Loop to check the market each morning and trade when specified conditions are met. Robinhood also describes overnight strategies that run continuously. Those examples and Loop capabilities

Fortune reported that a demonstration offered OpenAI’s GPT-6 Luna and GPT-6 Sol, alongside Anthropic’s Opus 4.8. Separately, Investing.com reports that the agents can currently place long equity, options, and cryptocurrency orders. That describes order categories, not evidence that agents can pursue every strategy or trade every instrument. Reported model choices and supported order categories

Controls limit authority, not investment risk

Robinhood says each agent operates through a dedicated trading account. Users can set limits on how much it trades at a time and require final approval before an order goes through. Those controls give users ways to bound or interrupt delegated activity, but they do not establish that a strategy will perform well or prevent losses. The account and approval controls

That distinction matters because a Loop can keep checking markets and acting without a person initiating each individual task. A scheduled trigger makes the agent useful for recurring rules, while also making the quality of its instructions and the limits on its authority consequential.

Adoption and operating costs remain open questions

Robinhood said more than 150,000 customers had opened agentic accounts through its earlier, more technical tool, and that agents were transacting on its platform nearly 30 million times a day as of late September. The first figure refers to accounts opened through the earlier tool; the second is the company’s reported platform activity, not a measure of profitable trades or adoption of the new in-app service. Robinhood’s reported figures and their timing

For the rollout, Robinhood planned to offer GPT-6 Luna free through the end of 2026 and charge standard token rates for its OpenAI and Anthropic agents. Executives described typical transaction costs as negligible, but research-heavy use or rising compute costs could alter the bill. The unresolved product question is not simply whether a model can place an order: it is how much autonomy users will grant, and what those recurring decisions cost to supervise and run. The rollout pricing plan and cost expectations