More than 10 companies are developing cryptocurrency wallets for AI agents, betting that automated software will create a new category of high-frequency payments that card networks are poorly equipped to handle. The immediate market is crypto-native trading and data services, but the longer-term ambition is to make wallets the payment layer for software that can spend small amounts repeatedly while carrying out a user’s instructions.
The economic case rests on transaction size and frequency. An AI agent searching the web, purchasing data, executing a trade, or coordinating a task across several services could make hundreds or thousands of payments per day. The report estimates that individual actions could cost as little as $0.001 and, in extreme cases, $0.00001. Fixed card-processing fees and chargeback systems make such tiny transfers impractical through conventional payment rails.
A single user request could therefore generate 20 to 30 separate machine-to-machine payments as an agent buys information, calls paid APIs, or pays for computing resources. Wallet providers are positioning themselves to manage those flows, including signing transactions, setting spending permissions, and recording an agent’s payment history.
Trading remains the clearest early market
The most visible current use of autonomous wallets remains trading bots operating in cryptocurrency and prediction markets rather than consumer purchases. The report points to a widely followed Polymarket experiment earlier this year in which an AI agent received $50 to trade independently and was required to stop if its returns could not cover API and server costs. The agent completed trades successfully, and comparable agent-led trading systems subsequently appeared.
Such tests are small compared with established trading volumes, yet they offer a practical measure of whether an agent can pay for the digital services it consumes. An agent that can earn more from trades, research, or task execution than it spends on data and infrastructure has a clearer path to operating without continuous human funding.
LayerHub, a research group cited in the report, said automated programs accounted for more than 30% of active wallets on major prediction platforms by mid-year. LayerHub also reported that these programs posted profits at a 37% rate, compared with win rates of 7% to 13% for human users competing in the same environments.
The report also cites Minarsch, whose full identity and firm were not specified in the supplied material, as saying a custom bot made more than 4,000 trades in one month and achieved returns of 376% on “exact” bets. It describes another automated wallet that reportedly turned a $300 deposit into $400,000 by identifying short-lived differences between network prices and prediction-market odds. Those examples illustrate the type of activity agent wallets are being built to support, though isolated trading results do not establish performance expectations for other bots.
For human traders, the spread of automated systems changes the nature of short-term competition. Bots can monitor prices, interpret structured data, and submit transactions in milliseconds, making raw execution speed a weaker edge for manually operated strategies. Specialized sector knowledge, contract interpretation, and tools that track proven on-chain wallets may offer more durable advantages than reacting to headlines after automated systems have already acted.
Revenue models depend heavily on usage assumptions
The report models the potential revenue impact of agent-wallet adoption using Coinbase’s 9.2 million monthly transacting users, rather than the company’s roughly 120 million registered accounts. Its estimates depend on three inputs: the share of users who adopt agents, the number of agents used by each person, and the frequency of daily payment calls.
Under its conservative scenario, 10% adoption, one agent per user, and 50 daily calls would add an estimated $84 million in annual revenue, equal to a 1.2% increase. A neutral scenario assumes 50% adoption, two agents per user, and 200 daily calls, producing a projected $3.36 billion in annual revenue, or a 46.8% increase.
The aggressive case assumes every monthly transacting user employs three agents making 1,000 calls per day. That framework generates an estimate of $50.37 billion in annual revenue, roughly seven times Coinbase’s current total revenue according to the report.
The gap between those outcomes shows how speculative the revenue forecasts remain. Raising adoption from 10% to 100% is a tenfold increase, but the report’s combination of more agents and much higher transaction frequency expands the projected revenue difference to about 600 times, from $84 million to $50.37 billion. The model describes the scale that machine payments could reach if usage becomes routine; it does not show that consumers or businesses have adopted agents at those rates.
Wallet data could support agent financing
Wallet developers also see potential value in transaction histories beyond payment fees. If an agent receives verifiable income for completing tasks, its on-chain cash flows could help establish whether it can service borrowing. That could eventually support revenue-based financing, where capital is advanced against future income rather than traditional credit scores.
The report compares this possibility with Stripe Capital, launched in September 2019. Stripe used sales data from its payment network to assess merchant eligibility and credit limits, reducing reliance on external credit bureaus and extensive application paperwork. An agent wallet could offer an analogous record of revenues, spending, and repayment capacity if the agent becomes an identifiable economic participant.
That condition remains distant for most current bots. Trading agents and payment executors may move funds rapidly, but a wallet’s transaction record only becomes useful for financing when income is reliable, attributable, and sufficient to repay a loan. The transition from software that spends a user’s funds to software that independently earns recurring revenue is therefore central to the more ambitious lending thesis.
Technical and regulatory gaps remain
Several obstacles could limit deployment outside controlled crypto settings. AI agents can hallucinate, meaning they may generate incorrect instructions or misunderstand information before initiating a payment. Fraud-detection systems may also block automated behavior that resembles suspicious account activity, even when a transaction is authorized by the user.
The payments stack is fragmented as well. Protocols including x402, AP2, and MPP are pursuing different approaches to agent payments, creating compatibility questions for wallet builders, merchants, and application developers. A common framework for identity, permissions, payment limits, and dispute handling has yet to emerge.
Legal status presents another complication. AI agents are not legal persons, leaving unresolved questions around know-your-customer checks, liability, and compliance when an autonomous program controls or directs transactions. Wallet providers can place limits around an agent’s authority, but financial rules generally assume an accountable human or company sits behind an account.
Solana currently accounts for 49% of bot transfer volume across the open internet, according to the report, which also says the network processed 35 million machine payments by March. That concentration gives Solana an early operational role in agent-payment experiments, while also showing that the sector remains concentrated in a small number of crypto-native networks and applications.
The report compares the current stage of agent wallets with longer platform-building cycles, noting that Apple’s App Store took 15 years to reach a $10 billion annual commission market and that WeChat Pay took seven years to develop its large mini-program ecosystem. Agent wallets may follow a similarly extended path, with their near-term value tied less to mass consumer checkout and more to building transaction histories, payment permissions, and reliable rails for software that can act economically on its own.
Turn theory into practice: explore automated agent trading with Toobit’s AI Agent TradeKit for real-time crypto strategies.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
