Revolut Engineers Build AI-Driven Trading Desk in 30 Minutes, Raising Questions About the Future of Broker Platforms

A small internal experiment at Revolut is sparking a wider discussion across the retail trading industry. Engineers working on the company’s crypto exchange say they assembled a functioning AI-powered market-making system in roughly half an hour—using a conversational prompt rather than traditional development workflows.

The system was built by connecting Claude AI, developed by Anthropic, to the Revolut X trading API through the Model Context Protocol (MCP). Instead of writing extensive code or building custom integrations, the engineers relied on MCP to allow the AI model to interact directly with trading infrastructure.

The demonstration is prompting a broader question: if a trading workflow can be orchestrated through a simple prompt, how valuable are traditional broker platforms built around dashboards and interfaces?

A Quick Experiment That Challenged Product Assumptions

The project began as an informal side experiment by two engineers on Revolut’s crypto team, Nikita Ivanov and Vlad Kaminski. According to Leonid Bashlykov, Head of Crypto Product at Revolut, the integration was initially developed simply to test Claude’s automation capabilities. Posting about the project on LinkedIn, Bashlykov said the results quickly exceeded expectations.

After connecting Claude to Revolut X through MCP, Bashlykov said he was able to build a working market-making workflow in around 30 minutes without prior preparation. The setup handled several tasks that typically require multiple systems, including portfolio screening, trade execution, monitoring, position management, and automated alerts.

The experiment, he said, forced the team to rethink assumptions about the future direction of trading products.

What the Model Context Protocol Enables

MCP, developed by Anthropic, is an open standard designed to allow AI models to discover and interact with external tools through a unified interface.

Traditionally, connecting a trading API to automation software requires custom-built integrations. MCP changes that by creating a standardized layer that AI systems can use to communicate with compatible services. This means an AI model can simultaneously access multiple resources—such as trading APIs, market data feeds, analytics tools, and messaging systems—without requiring separate engineering work for each integration.

Revolut X is currently offering MCP connectivity in beta, allowing developers and internal teams to experiment with AI-driven workflows.

From Platforms to Prompts

One of the most notable implications of the experiment is the shift in how trading instructions can be delivered.

Rather than navigating trading dashboards or writing algorithmic strategies, users could theoretically describe trading conditions in plain language. A prompt such as “rebalance the portfolio if Bitcoin dominance drops below a certain level” could trigger a full chain of automated actions.

Fintech analyst Linas Beliūnas highlighted the development in his Weekly Fintech Pulse newsletter, suggesting that MCP-based systems could represent a significant architectural shift from traditional trading bots. Conventional bots operate by executing predefined instructions. By contrast, AI agents can gather context, evaluate market conditions, and determine how to respond dynamically. According to Beliūnas, the type of workflow demonstrated by Revolut—covering inventory management, execution, monitoring, and notifications—would normally require significant technical infrastructure and weeks of development.

Why Revolut May Be Positioned to Move Faster

Revolut’s ability to experiment with AI integrations is partly the result of its technology architecture. The fintech company has gradually expanded its trading ecosystem through an API-centric strategy. In 2024, it integrated contract-for-difference trading through a partnership with CMC Markets’ institutional division, while a separate agreement with GTN added bond trading capabilities for European customers. This API-first approach makes it easier for external systems—including AI models—to interact with trading infrastructure directly. Brokers whose technology is tightly tied to proprietary front-end platforms may face greater challenges adapting to AI-driven workflows.

Rethinking the Broker Interface

The implications could extend beyond engineering experiments.

For more than a decade, retail brokers have competed by improving trading interfaces—offering faster order execution, advanced charting tools, and more intuitive mobile apps. But if AI agents become capable of executing complex trading strategies based on conversational prompts, the interface itself may become less central to the user experience. Instead, the competitive advantage could shift toward firms that build flexible APIs and data infrastructure capable of supporting AI-driven systems.

What began as a 30-minute experiment inside Revolut’s crypto team may therefore highlight a broader transition underway in financial technology: the move from trading platforms designed for screens to systems designed for machines.

Related read: The push toward faster and more flexible trading infrastructure is also visible elsewhere in the industry. In our article “Interactive Brokers Enables Near-Instant Trading via Stablecoin Deposits,” we report how Interactive Brokers introduced a system allowing clients to fund trading accounts using stablecoins, enabling near-instant deposits and quicker access to markets.