We break down Coinbase’s new tool, Coinbase for Agents, which allows AI models to trade autonomously while humans take a back seat.
On June 12, 2026, Coinbase unveiled Coinbase for Agents. The tool lets AI models like ChatGPT and Claude connect to a user’s account and execute trades and payments independently. The exchange also launched Coinbase Advisor, a registered AI agent built directly into the app that offers trading recommendations.
Hot topic: Bitcoin Mining Machine Profits Squeezed as Difficulty Cut Nears
The new tool gives AI agents access to spot and derivative crypto trading within user-set limits. Agents can rebalance portfolios, place limit orders, set up recurring buys, and pay for data using the x402 payment protocol. Each session runs in an isolated portfolio with no access to the user’s other assets.
Coinbase emphasizes that all transactions go through standard KYT checks and transaction monitoring. Users can set maximum trade sizes and spending limits. The tool is available through the Model Context Protocol for AI web interfaces and via CLI for developers.
Read more: Top 5 AI Crypto Coins to Watch in June 2026
Coinbase Advisor and Other Plans
Coinbase Advisor is an investment adviser registered with the SEC and CFTC/NFA and integrated directly into the exchange’s app. It analyzes the market and offers trade recommendations without any external setup.
The company says the tool lets users manage assets “without constant manual oversight.” For example, an agent could analyze historical data and automatically buy Ethereum (ETH) at the optimal price within a specific time window.
Coinbase plans to soon expand the agents’ capabilities to include stock trading, index funds, prediction markets, and commodities.
Read more: AI vs Crypto Smart Contracts — Can AI Replace Code Auditors?
Coinbase for Agents: Are There Risks for Users?
The launch reflects a growing trend of AI agent integration into crypto. Circle and Crossmint have released similar tools. According to Keyrock, AI agents conducted 176 million transactions worth $73M between May 2025 and April 2026.
But research paints a mixed picture. Studies from Pantera Capital and Stanford show that many agents don’t yet demonstrate fully autonomous trading. On average, users may end up with losses. Experts stress the need for strict limits and user oversight.
Learn more: How AI Agents Trade Crypto — Autonomous Trading Bots Explained
