The most API-native asset class on Robinhood is the one its trading robots still cannot touch. Robinhood Markets launched Agentic Trading in May 2026 for equities and options only, and by the end of the second quarter the feature had attracted nearly 100,000 accounts holding more than $100 million in assets under custody — an average of roughly $1,000 per account, a figure that says the product is being tested, not yet trusted. Crypto support, Chief Executive Officer Vlad Tenev told analysts on the July 29, 2026 earnings call, is coming next — which means the brokerage built its artificial-intelligence trading rails around the asset class with market hours and settlement cycles, while the one that trades 24/7 through public Application Programming Interfaces (APIs) waits in the queue.
Agentic trading is growing faster than the assets inside it
The headline numbers came alongside a record quarter. Robinhood reported $1.31 billion in second-quarter revenue against a $1.28 billion consensus, with earnings of $0.62 per share beating the $0.42 estimate, according to Benzinga’s July 30, 2026 earnings coverage. The stock still fell about 4% after the print, partly because crypto trading revenue cooled from the prior quarter, CoinDesk reported on July 29, 2026 — a reminder that the brokerage’s crypto franchise is cyclical while its product pipeline is not.
Tenev said the launch was deliberately conservative: agents were restricted to stocks at first, and users must fund a separate agentic account rather than link their primary balance. Even so, adoption crossed six figures within a single quarter — faster uptake than most of the firm’s product launches — while the custody base stayed small.
The models “fight you,” and that is the interesting part
The friction is not demand; it is the models and the plumbing. “Not everyone loves, surprisingly, going to a Codex or a Claude code and kind of stitching together these two apps,” Tenev said on the call, describing the technical sophistication the current setup demands. He added that the underlying AI models are not accustomed to being used for trading and will sometimes resist executing orders — behaviour Benzinga summarised in Tenev’s words as the models “fight you.” For institutional readers, that resistance is worth reading as an unpriced safety layer: the refusal behaviour trained into general-purpose models is currently doing risk-management work that no exchange rulebook has yet had to codify.
Exchanges are already racing for the same order flow
The venue side is not standing still. Kraken has been building institutional rails since it launched a prime brokerage platform to compete with Coinbase and FalconX, and the same infrastructure logic applies to agent-originated flow: whoever offers the cleanest programmatic access captures the order flow when retail agents finally connect to crypto venues. Robinhood’s own economics make the stakes explicit. The firm still earns a large share of revenue from routing arrangements, and as The Industry Spread has reported, the regulatory treatment of that model now splits three ways between the EU, UK and US after the June 30 payment-for-order-flow cliff, with Europe’s MiFIR Article 39a ban closing the carve-out Germany once enjoyed. An AI agent that trades around the clock generates exactly the kind of high-frequency retail flow that makes those routing questions commercially material again.
What happens when the agents get crypto keys
Tenev has argued publicly that AI agents could eventually trade with the capability of humans, telling CNBC on July 2, 2026 that agent-driven markets are a matter of when, not if. The near-term test is narrower: whether Robinhood ships crypto support for Agentic Trading while the accounts are still measured in the hundreds of thousands, and whether custodians and exchanges treat agent-originated orders differently for surveillance and accountability purposes. Analysis published August 1, 2026 framed crypto as the natural home for agentic flow precisely because the venues never close. If the integration lands in Q3, the metric to watch is not account count but assets per account — the moment that $1,000 average starts climbing, agentic trading stops being an experiment and starts being order flow.
This article is informational analysis only and is not financial, investment, or trading advice. Cryptocurrencies are highly volatile and can lose substantial value rapidly. Past performance and historical patterns do not guarantee future results. Do your own research and consult a regulated financial adviser before making any investment decision.
Karthik Subramanian
Karthik Subramanian is a founder, writer, and technology consultant with nine years in the crypto ecosystem. He covers token economics, L1/L2 infrastructure, DeFi protocols, wallets/custody, and the bridge between crypto and forex—broker technology, liquidity, and macro drivers. Karthik’s writing focuses on clear, practical frameworks that help professionals evaluate new products and on-chain innovation alongside FX market realities.
