What happened
Coinbase introduced Coinbase for Agents, a product that gives AI agents authenticated access to a user's Coinbase account so they can execute trades, send payments, and rebalance holdings autonomously. Per CoinTelegraph's report Thursday, the company pitched the tool as a way to manage crypto positions "without the constant manual oversight" that retail users currently sign up for. The announcement did not detail a public fee schedule, but framed the offering as available to both individual customers and developers building agent-driven applications on top of Coinbase's APIs. It is the most aggressive move yet by a US-listed exchange to embed large language models directly into the order path.
Under the hood, the product extends Coinbase's existing developer stack with agent-specific authentication and what the exchange describes as scoped permissions, so a user can in theory cap what an agent can spend, where it can route funds, and which assets it can touch. The exchange has not yet published the full permissioning matrix. Coinbase CEO Brian Armstrong has spent the past 18 months pushing the firm toward what he calls "machine-to-machine" payments, and this is the first product that puts that vocabulary into a customer-facing release.
Why it matters
This is the first time a top-five global exchange by spot volume has shipped a production tool that hands trading authority to an LLM. Until now, agentic crypto plays have been hobbyist projects, niche wallet integrations, or research demos. Coinbase has 8 million-plus monthly transacting users and roughly $400 billion in custodied assets as of its last 10-Q, which means even a single-digit-percent adoption rate puts meaningful volume in the hands of software that is not a human.