.png){kind=logo}
Coinbase has launched Coinbase for Agents, a new platform that lets artificial intelligence agents connect to user accounts and carry out crypto-related tasks. The product allows AI assistants such as ChatGPT-style and Claude-style agents to trade crypto, access market data, send payments and manage basic account workflows under user control.
The launch places Coinbase deeper into agentic finance, where AI tools can act on financial instructions without constant manual input.
Coinbase for Agents allows users to give AI agents access to trading and payment tools through natural language commands. A user could ask an agent to rebalance a crypto portfolio, monitor positions, identify market setups, or execute a trading plan within approved limits.
According to reports, the tool went live on Wednesday and supports crypto trading, account-linked workflows, and access to market data. Coinbase also plans to expand the service to stocks and prediction markets, although those features are not part of the full launch yet.
The company says agents can operate within user-defined controls. These include isolated portfolios and planned limits on spending, trading, and access to outside services. Nevertheless, early details on model support, approval settings, and launch-stage guardrails vary across reports.
Coinbase is also connecting the product to x402, its machine-to-machine payments protocol. The system allows AI agents to pay for digital services such as research, data APIs, and on-demand computing without using standard subscriptions or manual checkout steps.
This means an agent could gather paid market data, process that information, and then act on a user-approved trading instruction. Meanwhile, Coinbase is positioning x402 as part of a broader shift toward agentic commerce, where AI tools handle online payments and purchases.
Coinbase’s AI product lead, Lincoln Murr, said, “The whole idea is to give agents access to money and, through that financial independence, improve their set of capabilities to pretty much anything on the internet.” The quote reflects Coinbase’s goal, though adoption will depend on user trust and account protections.
The launch also brings risks tied to automated money movement. When software can trade or send payments, users need clear controls over what an agent can do, how much it can spend, and which services it can access.
However, questions remain around prompt errors, account misuse, and compromised agents. A wrong instruction or malicious service could create losses if permissions are too broad. For that reason, security controls will be central to how users assess the product.
Coinbase said agents can work inside isolated portfolios and will support customizable safeguards. These controls are expected to include spending caps, trade limits, and access restrictions. Additionally, audit trails and strong authentication may help users track actions taken by agents.
Coinbase is not the only company testing AI-led trading tools. Robinhood announced a similar product last month, showing that major trading platforms are moving toward AI-assisted account activity.
The trend comes as retail finance platforms seek faster and simpler user workflows. Crypto traders often respond to market shifts quickly and automated agents could reduce the need to manage every step manually.
Still, the model depends on user consent and clear limits. Coinbase for Agents gives AI tools a larger financial role, but its use will depend on how well the platform balances automation, account access and user control.
