Coinbase launches Coinbase for Agents, letting AI models make payments and trade crypto

Coinbase has rolled out a tool that lets AI agents connect to user exchange accounts, place trades, and make payments without human intervention. For high-risk operators, the useful part is not the AI angle; it is the plumbing: autonomous agents now have a payments rail and a trading interface inside a regulated exchange environment.

1. Coinbase said Thursday that Coinbase for Agents allows AI models such as ChatGPT and Claude to connect with a user’s exchange account through a model context protocol (MCP), so the model can be prompted to make trades or execute strategies. The same tool is also available through a command-line interface for developers.
2. The company said AI agents can use Coinbase’s AI payments protocol x402 to make payments, including paying for data services needed to collect information for trading strategies. In other words, the bot can fetch inputs, pay for them, and continue acting without a human clicking through each step.
3. Coinbase also introduced Coinbase Advisor, an AI agent integrated into its app that it says is a US Securities and Exchange Commission- and Commodity Futures Trading Commission-registered financial adviser. Coinbase said it can provide guidance on trades and help users manage crypto “without the constant manual oversight,” including allocating funds to reward programs or making recurring buys.
4. The company gave a concrete example: if a user wants to dollar-cost average into ETH at the optimal time of day, the agent can pull 30 days of hourly price data, identify when ETH historically trades lowest, set a recurring $20 market buy at that time, and schedule it to run daily for the next two weeks.
5. The launch lands as more crypto firms pitch AI agents as a way to handle high-frequency microtransactions. Coinbase is joining Circle, which last month launched tools for AI agents to use wallets, discover services, and make programmable payments with its token. Jeremy Allaire, Circle’s CEO, has predicted that billions of AI agents will use stablecoins within five years.

There is also a less polished side to the story. A study published last month by researchers at Pantera Capital, Stanford University, Ava Labs and the Initiative for Cryptocurrencies and Contracts found that, among over 925,000 token holders, agent treasuries made gains of $30 million on paper while token holders collectively lost $191.7 million. The same study said many projects did not yet provide clear evidence of autonomous trade execution, with a substantial share amounting to basic API integrations.