Gemini launches Agentic Trading for AI-powered crypto trading

Gemini rolled out Agentic Trading on Monday, a feature that lets users link AI models such as Claude and ChatGPT to their trading accounts on the exchange. The models can monitor markets, place and execute orders, and apply risk limits through Gemini’s U.S.-regulated trading API.

The feature integrates Gemini’s full trading API with the MCP open standard, a protocol developed to connect AI agents with external tools and application programming interfaces. Users can authorize agents to run continuously, identify market patterns, execute trades and enforce predefined risk settings without manual intervention.

Agentic Trading includes pre-built modular functions called Trading Skills. Examples include Find the Spread, which queries the bid-ask spread for any trading pair, and Retrieve Candles, which returns historical price data for pattern recognition and backtesting. Gemini said it plans to add more functions over time.

Gemini called the product “the first agentic trading tool to be available directly through a regulated US-based exchange.” The company added, “Agentic trading isn't just a feature. It's a new paradigm where AI handles the execution, patterns, and discipline, while you focus on strategy and goals.”

The release joins broader industry efforts on protocols that give AI agents automated access to wallets, payment rails and services. One project, x402, was incubated by another exchange and is now overseen by the Linux Foundation. Developers are also building Machine Payments Protocol standards. Both x402 and some machine payments efforts make use of MCP; Gemini’s tool is explicitly focused on executing trades on an exchange.

Gemini went public late last year. The company has recently seen departures in its finance and legal leadership, including its CFO, COO and CLO. A financial firm reduced its price target for GEMI, citing weak trading volume. At publication, Gemini shares traded around $4.45, down from their $28 debut in September.

On-chain activity this year included a March transfer of $130 million in bitcoin by the exchange founders to Gemini hot wallets. Agentic Trading is optional and operates only with user consent and the permissions a user grants to AI agents through the exchange API. Gemini noted the feature runs through its standard trading API and relies on user-defined parameters for strategy and risk management.

Questions remain about how automated agents will be monitored and how risk controls will be enforced on regulated platforms. Gemini said Agentic Trading operates within its existing API controls and the permissions framework users set for connected AI agents.