Robinhood lets AI agents trade stocks and use credit cards

Robinhood has launched a platform that lets users connect AI agents to their brokerage accounts to execute stock trades autonomously. The company is also introducing a virtual credit card designed specifically for AI agents, making it the first major consumer brokerage to open both trading and spending to autonomous software.

The agentic trading system works through a dedicated account structure. Users create a separate trading account for their AI agent, load it with a specific balance, and connect the agent to Robinhood’s Model Context Protocol service. The agent can then analyse portfolios, assess concentration risk and sector exposure, review analyst notes, and execute trades, but only using the funds in its dedicated wallet.

The safeguards are deliberate. Agents cannot access a user’s full brokerage balance. Users receive notifications for every trade an agent makes and can monitor activity within the Robinhood app. For certain trades, agents must show a preview and wait for user approval before executing. Robinhood said it has also built fraud detection into the system, with a team that reviews suspicious trades and helps resolve disputes.

The feature is launching in beta and is limited to stock trading. Robinhood said it plans to add options, crypto, event contracts, futures, and prediction markets in future updates.

The virtual credit card is a separate but related product. Gold Card holders can link their account to a new virtual card number, distinct from their primary card, that an AI agent can use to make purchases. Agents using the card earn 3% cashback, the same rate as the standard Gold Card. Users can set monthly spending limits and choose whether the agent needs approval for every transaction or only for purchases above a certain amount. The virtual card can be deleted at any time.

Robinhood’s Platinum Card, a $695-per-year premium card launched by invitation in March 2026, will get a similar agentic card feature when the product reaches general availability later this year.

Abhishek Fatehpuria, Robinhood’s vice president of product, told TechCrunch that the company built the feature in response to customer demand. Users wanted to bring their own tools, large language models, and agents and connect them to the platform. Fatehpuria, who joined Robinhood as an intern in 2016 and now leads the brokerage product, framed the launch as meeting early adopters where they already are.

The timing positions Robinhood in the middle of a broader race to build the financial infrastructure for AI agents. Stripe has launched an Agentic Commerce Suite with Shared Payment Tokens, a new payment primitive that lets AI agents initiate purchases using a buyer’s saved payment method without exposing card credentials. Amazon Web Services introduced Bedrock AgentCore Payments in partnership with Coinbase and Stripe, enabling AI agents to make stablecoin-based purchases. Google announced Universal Cart at I/O 2026, with an updated Agent Payments Protocol that lets AI agents complete purchases across Search, Gemini, and YouTube.

Visa and Mastercard are also moving. Mastercard launched Agent Pay in 2025, introducing agentic tokens that let AI agents initiate payments within defined boundaries. Visa partnered with Ramp on a trusted agent protocol for corporate expense management. Meow Technologies launched what it called the first agentic banking platform earlier this year, targeting the same gap between AI capabilities and financial infrastructure.

What makes Robinhood’s move distinctive is that it is handing AI agents the ability to trade securities, not just make purchases. That introduces a different category of risk. Stock trading involves market timing, price sensitivity, and regulatory obligations that do not apply to buying a pair of shoes. FINRA Rule 3110 requires firms to maintain human oversight over trading activity, and the SEC’s Market Access Rule mandates pre-trade risk checks. Whether an AI agent making autonomous trades on behalf of a retail investor satisfies those requirements is an open question.

Robinhood has navigated regulatory scrutiny before. The company paid $65 million to settle SEC charges in 2020 over misleading customers about how it generated revenue, and FINRA fined it $70 million in 2021 for systemic supervisory failures. Adding autonomous AI trading to a platform already under regulatory attention is a calculated bet that the controls, dedicated wallets, notifications, trade previews, fraud detection, are sufficient to satisfy regulators.

The company’s broader trajectory supports the gamble. Robinhood reported $4.5 billion in revenue in 2025 and $1.07 billion in the first quarter of 2026, up 15% year over year. It has 4.3 million Gold subscribers, a 36% increase from the prior year. Its market capitalisation, which hit $101 billion in December 2025, has since pulled back to roughly $62 billion. The company is no longer the volatile meme-stock platform of 2021. It is building itself into a full-service financial platform, and agentic trading is the next piece.

The acquisition of AI research platform Pluto in July 2024 laid the groundwork. Pluto’s technology uses large language models to analyse market data and generate investment strategies, and founder Jacob Sansbury joined Robinhood to lead AI integration. Last year, the company added an AI assistant that offers investment analysis and portfolio insights. The agentic trading launch extends that line from AI-assisted to AI-autonomous.

The question is whether retail investors are ready for it. Algorithmic trading has been standard in institutional finance for decades, but giving individual users the ability to hand a portfolio to an AI agent is a different proposition. The dedicated wallet structure limits downside exposure, but an agent making rapid trades on a volatile stock could still burn through a pre-loaded balance quickly. Robinhood’s beta framing suggests the company expects to learn from early usage before widening access.

McKinsey projects that AI agents could be responsible for $1 trillion in US transactions by 2030. Robinhood is betting that a meaningful share of that will flow through trading platforms, not just shopping carts. Whether regulators, and retail investors, agree is the part of the experiment that has not been written yet.