Crypto exchange Gemini rolled out Agentic Trading on Monday, becoming the first regulated US-based exchange to let artificial intelligence models connect directly to user trading accounts and execute orders without manual intervention.
The feature runs on the Model Context Protocol (MCP), an open standard originally built by Anthropic and now housed under the Linux Foundation’s Agentic AI Foundation. Users can link AI models from Anthropic’s Claude and OpenAI’s ChatGPT to their Gemini accounts, define strategies in plain language, and hand over execution to the bot.
How It Works
Agentic Trading turns Gemini’s full trading API into something an AI model can use natively. Once connected through MCP, the agent can scan markets, read order books, pull candlestick data, and place trades on the user’s behalf.
Tyler Winklevoss posted on X: “Calling all robots. Check out Gemini’s new MCP API, designed specifically for AI agents to trade predictions, crypto, and more.”
The system ships with pre-built modules Gemini calls “Trading Skills.” These are plug-and-play functions the AI agent can call mid-session:
- Find the Spread queries bid-ask data for any pair listed on the exchange
- Retrieve Candles pulls historical price charts for pattern analysis and backtesting
The modular setup means Gemini can keep stacking new skills without tearing apart the core architecture.
Human Guardrails Stay in Place
Gemini was careful to frame this as a tool, not a free-for-all. Users set the parameters. The AI operates inside those guardrails. You decide the risk tolerance, the pairs, the position sizes. The agent handles timing and execution.
That distinction matters in a market where regulators watch every product launch from crypto firms. Gemini holds a New York BitLicense and operates as a regulated exchange – pushing out autonomous trading bots without safety rails would have invited scrutiny from the DFS and SEC.
Why MCP Changes the Game
Before MCP, hooking an AI model up to an exchange required custom wrappers, proprietary SDKs, and a developer who could wire everything together. MCP standardises that plumbing. Any model that speaks the protocol can plug into any service that supports it.
Anthropic shipped MCP in late 2024 as an open standard. Since then, OpenAI, Google DeepMind, and Microsoft have adopted it. The protocol was later donated to the Agentic AI Foundation, co-founded by Anthropic, Block, and OpenAI under the Linux Foundation umbrella.
For crypto, this is a big deal. It means a trader can swap from Claude to ChatGPT to whatever model ships next month without rebuilding their entire workflow. No vendor lock-in, no proprietary prison.
Coinbase Already Moving the Same Direction
Gemini isn’t operating in a vacuum. Coinbase launched its own AI-focused developer toolkit earlier this year, and several smaller exchanges have opened API access for bot-driven trading. But none had packaged the experience this cleanly for retail users on a regulated platform.
The distinction between infrastructure for developers and a consumer-ready product matters. Gemini is aiming at the trader who wants to say “buy BTC when RSI drops below 30 and funding rates flip negative” in plain English and walk away.
Market Context
The launch drops during a week where Bitcoin sits around $77,000 and the broader market is consolidating after April’s rally. Automated strategies tend to gain traction during range-bound periods when human traders get impatient and miss entries.
Gemini’s move also overlaps with growing interest in AI agents across crypto. Alchemy’s CEO recently stated that crypto infrastructure was “built for AI agents, not humans” – a provocative framing, but one that Gemini’s product now gives real weight to.
What Comes Next
Gemini said it plans to expand Trading Skills and deepen MCP integration over the coming months. The company didn’t provide specifics on whether agentic trading would extend to derivatives, prediction markets, or its custody products.
For now, the feature is live for all Gemini users. No waitlist, no beta gate.
FAQ
Can AI bots withdraw funds from my Gemini account?
No. Agentic Trading is scoped to trading execution only. Withdrawal permissions aren’t part of the MCP integration.
Which AI models work with Gemini Agentic Trading?
Currently Claude (Anthropic) and ChatGPT (OpenAI) are supported. Any MCP-compatible model should work .
Do I need coding experience to use it?
No. Gemini’s Trading Skills are pre-built modules you can configure in plain language through your AI model of choice.
