Robinhood is extending agentic trading to crypto, giving customers a way to let AI agents buy and sell digital assets while the market stays open through the night.
Robinhood is moving the AI trading experiment into the part of finance that never closes. At a July 1 event in London, the Menlo Park brokerage said U.S. customers will be able to connect their own AI agents to trade cryptocurrencies on their behalf, with no extra fee, according to Barron's.
That date matters. This wasn't a July 10 standalone launch, and it didn't arrive nine days after Robinhood Chain. The crypto trading expansion and the Robinhood Chain mainnet were part of the same product push. Get that wrong and you miss the point: Robinhood isn't treating AI agents, crypto and tokenized assets as separate experiments. It's putting them on the same rail.
This is a risky place to do it. Stocks stop trading. Crypto doesn't. You can go to bed with Bitcoin steady and wake up to a liquidation, an exchange outage, a weekend policy shock or a move that looks absurd until you remember this market has done it before. A trading agent doesn't need sleep. You do.
Robinhood first opened agentic trading for stocks in late May. Axios reported at the time that users could plug outside AI agents into Robinhood through its Model Context Protocol servers, create a separate agentic trading account with dedicated funds, and watch activity and profit and loss inside the Robinhood app. The Verge also noted Robinhood's warning that users could lose their entire investment. That's not fine print. That's the product.
The guardrail is the account #
The core safeguard is separation. Customers don't hand an agent their whole portfolio. They put money into a dedicated agentic account, then let the software trade only inside that account. Robinhood has said users receive alerts when agents trade and can disable agent access.
On paper, that's sensible. Frankly, it's also the minimum you would expect before letting software turn a prompt into an order. A contained account can still lose real money. A notification after the trade is useful, but it isn't the same as judgment before the trade. If an agent buys the wrong token at 3 a.m., your phone lighting up doesn't make the decision smarter.
The harder question is authorization. If you tell an agent to rebalance into crypto assets when momentum improves, did you authorize one trade, a strategy, or every mistaken interpretation that follows? Robinhood can build visible controls around the account. It can't make every customer understand how a model arrived at a trade, or whether the model understood the instruction at all.
That's why crypto is the sharp test. The market is open every hour, and the inputs are noisy: prices, social posts, token unlocks, wallet movements, protocol news, exchange listings, regulatory comments, you name it. Humans already overtrade that mess. An agent can do it faster.
Robinhood wants the whole stack #
The crypto agent feature also sits beside a broader blockchain push. Barron's reported that Robinhood launched the Robinhood Chain mainnet on July 1 to support real-world assets, while tokenized stocks on Robinhood Chain won't be available to U.S. customers. Eligible U.S. customers are instead getting Robinhood Earn, a stablecoin lending service using USDG through a self-custody wallet and infrastructure from Morpho.
That mix tells you where Robinhood is going. It wants the brokerage account, the AI interface, the crypto wallet, the tokenized asset layer and the lending product close together. Coinbase and Kraken should pay attention, but so should Charles Schwab and Interactive Brokers. Robinhood is trying to make the retail trading account feel less like a place you visit and more like software that acts while you're away.
Vlad Tenev has been making the same case in public. The Economic Times reported this week that Robinhood's chief executive said AI agents will soon rival human traders and manage portfolios. You don't have to accept that prediction to see the direction of travel. Robinhood is not pitching AI as a research assistant anymore. It's pitching AI as an actor.
That changes the customer's job. You are no longer just choosing whether to buy Bitcoin, Ether or a tokenized asset. You are choosing how much authority to give a system that can misunderstand you, act quickly and keep acting when you're not watching. The off switch helps. The better question is how often you'll know to use it in time.
Also read: LAB Token Collapses Nearly 98% After ZachXBT Traces Dump to Team Funded Wallet • Eric Trump's Ethereum Cheerleading Comes With an Undisclosed Financial Stake • Empery Digital Dumps Nearly Half Its Bitcoin to Chase an AI Data Center Bet