Over 70,000 traders have already handed their equities and options portfolios to AI agents on Robinhood since late May, and the brokerage said Friday that crypto users are next.
The company announced during a presentation that eligible US customers will “soon” be able to connect third-party AI agents to execute Bitcoin, Ethereum, and other cryptocurrency trades autonomously. No specific launch date was given, but UK customers were identified as next in the rollout queue after the US crypto expansion.
“You can work with an agent to create a strategy with specific guardrails and not need to be constantly monitoring your account,” a Robinhood executive said during Friday’s event. The framing positions autonomous trading as a time-saver for retail users who lack the bandwidth to watch charts all day, though it also creates doubt about what happens when an AI decides to sell during a flash crash while you’re asleep.
Equities Beta Shows Strong Early Adoption
The numbers from Robinhood’s equities beta suggest meaningful traction. More than 70,000 agentic accounts have been created since the late May launch, roughly seven weeks of activity. That works out to about 10,000 new agentic accounts per week on average, a pace that indicates retail appetite for delegation exists even before the feature carries any significant track record.
Robinhood’s AI agent infrastructure relies on partnerships rather than in-house models. The platform offers connections to Anthropic (maker of Claude), OpenAI (ChatGPT and GPT-4), and SpaceX’s Grok. Users pick their preferred AI provider and define trading parameters, letting the agent handle execution within those constraints.
The current equities product already touches crypto indirectly. Traders can ask AI agents to invest in crypto mining stocks on their behalf. That creates a scenario where an AI agent might buy Marathon Digital or Riot Platforms shares based on Bitcoin sentiment without the user ever touching the underlying asset. The upcoming crypto expansion removes that extra layer, putting agents directly in contact with spot markets.
The pitch is familiar: level the playing field. Institutions have run algorithmic and quantitative strategies for years, executing at speeds and scales individual traders cannot match. Whether a third-party AI agent with user-defined guardrails actually replicates that edge is a different question. The agent can monitor data streams the user would miss, sure, but it’s also operating with fewer resources, less proprietary data, and (presumably) less battle-tested risk management than a Renaissance Technologies quant desk.
Robinhood Chain Adds Infrastructure Context
The AI agent announcement lands a week after Robinhood launched its own Ethereum layer 2 network. Robinhood Chain went live earlier this month, and the early metrics are substantial: 17 million transactions from nearly 350,000 wallet addresses in the first week, according to Johann Kerbrat, Robinhood’s senior vice president and general manager of crypto.
Those figures suggest aggressive onboarding, though note that transaction counts can be inflated by bots, airdrops, or promotional activity common during chain launches. The 350,000 wallet count is a cleaner signal of distinct users, though even that metric doesn’t distinguish between real retail adoption and developers testing infrastructure.
Robinhood Chain is built on the Arbitrum stack and focuses on two pillars: real-world asset tokenization and yield products. The company is offering 7% APY on its USDG stablecoin through a lending product, a rate competitive enough to draw capital from traditional savings accounts and money-market funds. Tokenized equities are available to customers in over 120 countries.
The connection between the L2 launch and AI agents isn’t explicit in Robinhood’s announcements, but it’s not hard to imagine the pieces fitting together. An AI agent that can execute crypto trades might eventually be able to move assets onto Robinhood Chain, deposit into yield products, or trade tokenized stocks, all without human intervention. Whether regulators will be comfortable with that level of autonomy is another matter entirely.
Industry Momentum Behind Agent-Driven Payments
Robinhood isn’t operating in a vacuum. Multiple crypto executives have publicly predicted that AI agents will become the dominant users of blockchain payment rails within a few years. Coinbase CEO Brian Armstrong and Circle CEO Jeremy Allaire have both floated this thesis, envisioning a future where autonomous software pays for compute, bandwidth, and services in stablecoins without human approval for each transaction.
Several integrations have moved toward that vision. In May, Amazon Web Services integrated Coinbase’s x402 payments protocol into Amazon Bedrock AgentCore, letting AI agents transact in USDC. The integration is significant because it places stablecoin rails inside AWS’s massive cloud infrastructure, theoretically enabling any Bedrock-hosted agent to make USDC payments.
In April, crypto wallet startup Oobit launched a Visa-supported virtual card specifically for AI agents. The card allows agents to make online purchases in USDT on behalf of businesses. It’s a different use case from trading (more about expense management than speculation), but it reinforces the same underlying bet: machines will increasingly need native internet money.
The volume reality, however, is sobering. Artemis data shows that only $2 million in transaction volume was facilitated through the x402 protocol in June 2026. That’s essentially a rounding error relative to the broader stablecoin market, where daily volumes regularly exceed $50 billion. The infrastructure is being laid; the throughput isn’t there yet.
For Robinhood, the question is whether crypto trading volume from AI agents will follow a steeper adoption curve than agent payments for services. Trading has more immediate feedback loops (profit, loss, portfolio balance) than generalized spending, which might accelerate iteration and user trust. Or it might just accelerate losses if agents make poor decisions in volatile markets.
Risks and Unknowns in Autonomous Crypto Trading
The phrase “specific guardrails” from Robinhood’s presentation deserves scrutiny. What guardrails, exactly? A stop-loss percentage? A daily volume cap? A whitelist of tradeable assets? The company hasn’t published detailed documentation on how guardrails work, what defaults are applied, or how liability is allocated when an agent executes a trade the user didn’t anticipate.
Crypto markets present unique challenges for autonomous trading. They run 24/7, meaning an agent might execute during low-liquidity hours when spreads are wide and slippage is severe. Flash crashes happen faster than human reaction time, but they also happen faster than most users can wake up and override an agent’s decision. The Robinhood executive’s framing (“not need to be constantly monitoring your account”) cuts both ways: monitoring is exhausting, but it’s also a form of risk control.
There’s also the question of what happens to funding rates and market microstructure when thousands of AI agents are executing similar strategies simultaneously. If retail agents cluster around the same signals (because they’re using the same underlying models from Anthropic or OpenAI), they could amplify volatility rather than smooth it. Institutional quants have dealt with strategy crowding for years; retail AI agents are a new variable.
Regulatory clarity is another open question. The SEC and CFTC haven’t issued guidance specifically on AI-agent trading for crypto. Robinhood’s compliance team presumably vetted the rollout, but the regulatory landscape could shift. If an agent causes a significant loss and the user claims they didn’t understand what they authorized, where does liability fall? On Robinhood? On the AI provider? On the user who clicked “connect”?
The Bitcoin treasury strategies that have drawn institutional capital over the past two years involve human committees making allocation decisions. Autonomous agents represent a different model entirely, one where the machine executes and the human reviews after the fact. For some traders, that’s liberation. For others, it’s an invitation to lose money faster than they could manage manually.
What Comes Next
Robinhood’s crypto expansion arrives during a period of experimentation across the industry. Kraken announced plans to add an AI investing assistant to its app this week, taking a more advisory approach rather than full autonomy. Backpack is pursuing 24/7 stock markets through tokenized equities. The common thread is removing friction, human sleep schedules, manual order entry, market hours, and letting software do the work.
Whether that’s a feature or a bug depends on how much you trust the software. The 70,000 equities traders who signed up for Robinhood’s beta are placing an early bet that the upside outweighs the risk. Crypto traders will soon get to make the same choice, with the added wrinkle that they’re operating in markets where volatility can hit 10% in an hour and liquidity can evaporate without warning.
The democratization pitch is real, to a point. Retail traders have never had access to tools that execute while they’re offline, monitor data streams they’d miss, and react in milliseconds. But access to a tool doesn’t guarantee profitable use of that tool. The institutions Robinhood references have armies of risk managers, years of backtested strategies, and capital reserves to absorb drawdowns. A retail user with an AI agent and $5,000 has none of those buffers.
Robinhood didn’t disclose how the AI agents will handle crypto-specific considerations like network congestion, withdrawal limits, or on-chain settlement times. Those details will matter. A strategy that works on Nasdaq may not translate directly to markets where you can’t always get your money off an exchange instantly.
The company is clearly betting that retail demand for delegation is strong enough to justify the rollout. The beta numbers support that bet. What the numbers don’t show yet is whether the agents actually make money for their users, or whether they just make trading more convenient on the way to the same outcome.
