Coinbase Lets AI Agents Research Portfolios and Place Trades

Coinbase is offering AI-agent access through Coinbase Advanced, letting users connect assistants such as ChatGPT, Grok, Claude and Perplexity to research portfolios and place trades. Users can give instructions in natural language, approve trades themselves and choose which portfolio an agent can access.

The tutorial describes agents that can manage positions across crypto, equities and eligible futures, rebalance portfolios, track performance and run recurring strategies. A dedicated portfolio can limit an agent's access to funds transferred into it, rather than exposing a user's other holdings. Coinbase says there is no extra charge for the agent feature, though standard exchange fees still apply.

Why it matters

The setup brings conversational AI closer to trading execution, not just market research. Users can describe a strategy and have an agent monitor conditions or act on instructions, while avoiding the need to write code. That convenience also makes clear strategy rules and access limits important: vague instructions such as “make money” do not specify a reliable trading plan.

The tutorial emphasizes that AI agents can make mistakes and that users direct their activity and assume the risk of executed transactions. It recommends starting with a separate portfolio and a limited amount of funds, and cautions that users need to understand trading basics before delegating decisions.

Market impact

The feature does not establish that automated strategies will outperform manual trading. Its significance is the tighter connection between AI assistants and exchange functions, including trading across multiple asset classes where accounts are eligible. Coinbase says standard fees apply, and the tutorial also promotes a trading competition with a prize pool that can scale to $2 million in USDC as collective volume grows.

For traders, the practical test is whether an agent follows clearly defined rules and stays within its permitted portfolio. Automated execution can reduce friction, but it does not remove market risk or the need for user oversight.