Coinbase is extending its agentic trading platform beyond crypto and derivatives. This allows AI agents to trade U.S. stocks and ETFs while using micropayments to acquire data, model inference, and other services. The organization said Coinbase for Agents MCP now blends equities trading with its x402 payment protocol, allowing the agent to research an asset, pay for context, evaluate the results, and execute a trade within user-defined limits.
The extension is a part of Coinbase’s broader AiFi vision for financial systems that can operate recurrently through AI agents. The company is also positioning Coinbase for Agents as an “Everything Exchange” for AI agents, with ETFs, stock, crypto, and derivatives available through a single Coinbase login.
Coinbase Combines Stock Trading and AI Payments
Coinbase said agents can now trade fully owned U.S. stocks and ETFs through Coinbase for Agents, adding equities to the crypto and derivatives capabilities already available. The organization said the equities service uses Coinbase Capital Markets Corp., a FINRA/SIPC member, while Apex Fintech Solutions handles clearing and custody.
The platform gives agents access to thousands of U.S. stocks and ETFs, along with 24/5 trading, fractional shares starting at $1 and zero commissions. The company said the expansion essentially opens its current regulated equities business to a new type of user i.e. that is AI agents. Tokenized equities are also on Coinbase’s route for markets outside the United States.
The other essential component of the launch is x402, Coinbase’s payments protocol. Through x402, an AI agent can make small payments for context or model inference instead of needing users to maintain separate subscriptions, API keys or accounts. For example, an agent could pay a few cents for an earnings summary, an onchain analytics query or another model’s analysis. It can then use the information to determine whether conditions defined by the user have been met before taking another action.
Coinbase said the same balance can be used for both trading and these micropayments. This creates a single loop in which an agent can identify a probable indication, pay for information to confirm it and execute a trade. The company said x402 has processed more than 230 million transactions and more than $54 million in volume in just over a year. Coinbase also said it is the largest facilitator for the protocol, processing more than half of all x402 transactions.
The equities and micropayments combination expands what Coinbase sees as probable for agentic trading. An agent can monitor an asset continuously, obtain context when needed and act according to instructions instead of needing a user to supervise several platforms. Coinbase describes this broader approach as AiFi, or a financial system designed for trading that can work consistently.
AI Functioning with Coinbase’s New Trading Tools
Coinbase provided several examples of how it is testing the expanded capabilities. In one example, an employee named Sarah asked an agent to research NVDA, AAPL and GOOG ahead of an earnings cycle. The agent used x402 to pay for analyst sentiment and options-flow data , summarized the findings and then placed fractional purchases across the three stocks after receiving instructions to do so.
In another example, Lincoln, described as Coinbase’s AI product lead, configured an agent to supervise an ETH position. The agent checks onchain flow data and can identify a spike in exchange inflows. If the predefined conditions are met, it can execute a preset stop-loss without requiring Lincoln to supervise the position.
The example emphasizes the combination of automated monitoring, paid information access, and trading execution that Coinbase is introducing. Instead of an agent simply supervising a portfolio, it can obtain extra information when needed. It can also use that information as part of a trading workflow.
Coinbase highlighted that users continue to create the boundaries within which their agents can work. Users can restrict an agent to a specific portfolio, configure which assets it can access, build spending limits, and require approval for individual actions. Sandbox accounts for equities are also planned.
The company said setting up Coinbase for Agents can take as few as three clicks. Users can search for Coinbase through the Connectors or Plugins section in Perplexity or Grok and then sign in into their Coinbase account. Coinbase said users of ChatGPT, Claude, and other agents can connect through MCP.
The expansion means an AI agent can handle several parts of trading workflow from one single setup. It can monitor positions, research assets, pay for additional information, and execute trades. Coinbase’s approach is to keep the user in control through strategies and safeguards while allowing the agent to perform the recurrent work.
For Coinbase, the addition of stocks and ETFs represent another step in its effort to build an agent-focused financial platform. The company is exchanging the same infrastructure used for crypto trading and payment into a wider system where AI agents can interact with several asset classes and services through one single account.
