Gemini has launched a new feature called Agentic Trading, giving users a way to connect AI agents directly to their crypto trading accounts and automate market activity through the exchange’s own infrastructure. The company describes the product as the first agentic trading tool available directly through a regulated U.S.-based crypto exchange.
The launch marks a new step in the fast-moving overlap between artificial intelligence and digital assets. Instead of using AI only to summarize market data or generate trading ideas, Gemini’s new system allows AI tools such as ChatGPT, Claude, or other models to access exchange functions through the Model Context Protocol, commonly known as MCP. Once connected, an agent can monitor markets, read trading data, place orders, and help manage risk based on a user-defined strategy.
For crypto traders, the appeal is easy to understand. Digital asset markets never close. Bitcoin, Ethereum, stablecoins, and altcoins trade 24 hours a day, including weekends and holidays. A human trader can miss opportunities while sleeping, traveling, or working. An AI trading agent, at least in theory, can stay alert around the clock.
How Gemini Agentic Trading Works
The core idea behind Gemini Agentic Trading is simple: the user sets the strategy, and the AI agent handles execution. Gemini says agentic trading means an AI agent can act on a trader’s behalf by placing trades, watching markets, and managing risk automatically.
At the center of the product is MCP, an open standard originally introduced by Anthropic to create secure, two-way connections between AI-powered tools and external systems. Anthropic explains that MCP lets developers expose data through MCP servers or build AI applications that connect to those servers.
In Gemini’s case, MCP connects AI agents to Gemini’s trading API. The company says it has integrated its full API with MCP, meaning that actions previously available through a traditional API can now be performed by an AI agent. A trader can connect an AI model, define a strategy, and allow the agent to carry out instructions through Gemini’s exchange tools.
This does not mean the AI suddenly becomes a magic market predictor. It means the agent can interact with Gemini’s exchange system in a more natural, automated way. The trader still needs to decide what the strategy should be, what level of risk is acceptable, and how much capital should be placed under automated control.
Trading Skills Give AI Agents Market Tools
Gemini has also introduced what it calls Trading Skills, which are modular functions that AI agents can use when carrying out a strategy. At launch, those skills include pulling real-time market data, checking order book depth, finding bid-ask spreads, and retrieving historical candlestick data.
These tools are important because AI agents need reliable inputs before they can act. A model cannot responsibly execute a trading plan without access to current prices, liquidity conditions, spreads, and historical market behavior. Gemini’s prebuilt skills are designed to give agents those building blocks.
For example, a user might ask an agent to buy Bitcoin only if the bid-ask spread falls below a certain level, or sell part of a position if a price target is reached. Gemini’s own example says a trader could ask a model to sell BTC if Bitcoin reaches $100,000, or buy BTC when the bid-ask spread reaches 0.01%.
That kind of plain-language interaction could make automated crypto trading more accessible. Until now, algorithmic trading has often required coding knowledge, API experience, or third-party bot platforms. Gemini is trying to lower that barrier by letting users communicate strategy goals to AI tools they may already use.
A Shift From Trading Bots to AI Trading Agents
Crypto trading bots are not new. Traders have used bots for years to execute grid strategies, arbitrage trades, stop-loss systems, and portfolio rebalancing. What makes Gemini’s launch different is the agentic layer.
Traditional bots usually follow fixed rules. They do what they are programmed to do, and they often need manual configuration. AI agents can potentially interpret broader instructions, request data, evaluate market conditions, and choose which tools to use. That flexibility is what makes the new model powerful, but it is also what makes it more complex.
Gemini argues that agentic trading represents a broader shift in how people interact with financial markets. In the company’s framing, AI handles execution, pattern recognition, and discipline, while users focus more on strategy and goals.
That vision fits the wider AI industry trend. MCP documentation describes the protocol as an open-source standard for connecting AI applications to external systems, including data sources, tools, and workflows. In plain English, MCP gives AI models a more consistent way to use software outside the chat window.
In crypto, that outside software is an exchange account. That makes the stakes much higher than asking an AI assistant to summarize a document or draft an email. When an agent can place trades, mistakes can cost real money.
Why This Matters for Crypto Exchanges
Gemini’s launch also signals a competitive shift among crypto exchanges. Trading platforms have spent years competing on fees, liquidity, asset listings, custody, staking, derivatives, and user experience. AI-powered execution could become the next battleground.
If agentic trading becomes popular, users may expect exchanges to support AI-native workflows. Developers may build custom agents that connect to exchanges, analyze market data, rebalance portfolios, and automate strategies. More advanced traders could chain together multiple tools to build sophisticated systems without relying entirely on legacy trading bots.
Gemini says the product is designed for both experienced quant traders and users who are newer to automation. Advanced users can build custom agents and multi-leg strategies, while less technical users can start with prebuilt skills and simple instructions.
That flexibility could help the exchange attract developers and active traders. It may also deepen user engagement because traders who build automated systems around a platform are less likely to move elsewhere.
Risks Remain High for AI Crypto Trading
The launch also raises serious risk questions. Crypto markets are volatile, liquidity can change quickly, and automated strategies can fail when market conditions shift. AI systems may misunderstand instructions, rely on incomplete data, or act too quickly during sharp price swings.
U.S. regulators have already warned investors not to treat AI trading systems as guaranteed money machines. The Commodity Futures Trading Commission has cautioned that fraudsters use AI hype to promote automated trading algorithms and crypto-asset schemes promising unrealistic returns, adding that AI cannot predict sudden market changes with certainty.
The SEC, NASAA, and FINRA have also warned that bad actors are using the popularity of AI to promote investment fraud. Their investor alert urged people to be careful with claims involving artificial intelligence and emerging technology.
Those warnings are relevant even when the underlying exchange tool is legitimate. A regulated platform can provide infrastructure, but users still need to understand the risks of letting software trade on their behalf. No AI model can remove market risk, liquidity risk, execution risk, or the possibility of poor strategy design.
Security and Control Will Be Crucial
For agentic trading to succeed, security controls will matter as much as convenience. Users will need clear permissions, spending limits, order limits, and emergency shutoff options. Developers will need to think carefully about API keys, account access, model behavior, and prompt injection risks.
The more power an AI agent has, the more important guardrails become. A market-monitoring agent that only reads prices is low risk. An agent that can place live buy and sell orders carries a much higher level of responsibility.
Gemini’s move places this debate directly inside crypto trading. The industry has long embraced automation, but AI agents introduce a new kind of uncertainty because they can interpret instructions rather than simply follow rigid code. That makes usability better, but oversight more important.
What Comes Next for AI-Powered Crypto Trading
Gemini says it plans to expand its skills library, which could allow AI agents to perform a wider range of trading actions over time. If the product gains traction, other exchanges may respond with their own AI-agent integrations, developer tools, or MCP-compatible trading systems.
The bigger question is whether traders will trust AI agents with real capital. Some will see the technology as a productivity boost, especially for monitoring markets and automating simple rules. Others will stay cautious, preferring manual control in a market known for sudden crashes, liquidation cascades, and unpredictable news events.
Either way, Gemini’s Agentic Trading launch is an important moment for crypto market automation. It moves AI from the role of adviser to the role of active market participant. For traders, that could mean faster execution and more flexible strategies. For regulators and risk managers, it introduces a new layer of complexity.
The future of crypto trading may not be fully autonomous, but it is clearly becoming more automated. Gemini’s latest product shows that AI agents are no longer just watching the market. They are beginning to trade it.