Two engineers at Revolut’s crypto exchange built a fully
functional market-making system in about half an hour. Not a prototype. Not a
carefully prepared demo. Just a first idea, executed using Anthropic’s Claude
AI connected to the Revolut X trading API through a protocol called MCP.
The experiment is raising a question the retail trading
industry may not be ready to answer: if an AI agent can orchestrate an entire
trading workflow through a simple text prompt, what exactly is a broker’s
platform worth anymore?
Revolutโs Side Project That Broke the Product Roadmap
Nikita Ivanov and Vlad Kaminski, engineers on Revolut’s
crypto team, built the integration as a personal experiment, according to
Leonid Bashlykov, Revolut’s Head of Product for Crypto.
Leonid Bashlykov, Source: LinkedIn
“Two A-players on my team…were playing with Claude
automations and built the MCP as a side project, just for fun,” Bashlykov
wrote on LinkedIn. “Then we saw what it could actually do. And it broke
our product roadmap thinking.”
Bashlykov said he personally built a working market-making
strategy using Claude, MCP, and the Revolut X API in 30 minutes, with no
advance preparation. The workflow covered everything from portfolio screening
and position sizing to execution, monitoring and automated alerts. We are
talking about tasks that traditionally require separate tools, scripts, and a
fair amount of technical expertise to connect together.
MCP, or Model Context Protocol, is an open standard
developed by Anthropic that allows AI models to discover and use external tools
through a standardized interface. Rather than requiring custom-built
integrations for every API, MCP lets an AI agent connect to any compatible
system and reason across all of them simultaneously. Revolut X is currently
offering MCP access in beta.
Prompts Replace Dashboards
Linas Beliลซnas
The practical implication, according to Bashlykov, is that
complex trading strategies no longer require coding knowledge or platform
fluency. “Rebalance to 60/40 BTC/ETH if BTC dominance drops below
52%,” he wrote as an example. “That’s now a prompt.”
Fintech analyst Linas Beliลซnas, who covered the story in his
Weekly Fintech Pulse newsletter, framed it as a potential architectural shift.
“The agent reads portfolio data, checks conditions, pulls context – then
decides,” he wrote, contrasting MCP-connected agents with rule-based
trading bots that simply execute predefined instructions.
Beliลซnas noted that the workflow Bashlykov described,
handling inventory management, dynamic quoting, position sizing, execution,
monitoring and Telegram alerts, “normally requires a quant, a developer,
backtesting infrastructure, and weeks of iteration.”
Most AI trading tools currently on the market, both
Bashlykov and Beliลซnas agree, are still limited to price alerts, simple
queries, and basic automations.
Revolut engineers built an AI trading workflow with Claude in 30 minutes.
Now theyโre questioning whether they need to build product features at all ๐ค
Inside Revolut X, two engineers – Nikita Ivanov & Vlad Kaminski – started experimenting with Claude automations.
Just a sideโฆ pic.twitter.com/2YI65JygRv
โ Linas Beliลซnas (@linasbeliunas) March 4, 2026
What makes MCP different, the company claims, is
composability. A single connected workflow can simultaneously execute trades,
read news, track on-chain data flows and update dashboards. Revolut X functions
as the execution layer; Claude functions as the interface.
The Brokers Already Moving (and Those Who Aren’t)
Revolut is not entirely alone. CFD
broker ATFX partnered with data firm KX late last year to deploy an AI-driven
MCP server for real-time trading data, and LSEG connected its market data
feeds directly to ChatGPT in December 2025, allowing institutional users to
pull live information on demand. A FinanceMagnates.com analysis from November
2025 noted that crypto-native firms were already moving into AI’s data layer
while asking whether traditional brokers would follow.
The answer, so far, is: slowly. Firms like XTB, IG Group and
Plus500 have spent years and significant capital building proprietary trading
interfaces. XTB entered 2026 projecting $186 million in income while
doubling down on platform growth and client acquisition, a roadmap built around
the assumption that the interface is the product.
That assumption is now being tested. A comparison of client metrics across IG, CMC, Plus500 and XTB published
last August showed that client numbers are growing across the board, but
revenue per user varies dramatically. It signals that platform stickiness, not
just user count, drives broker economics. If an AI agent can replicate a
platform’s functionality through a plain-language prompt, the stickiness
argument weakens considerably.
Revolut’s Head Start
Revolut’s position in this shift is not accidental. The
company has systematically built API-first infrastructure across asset classes.
A partnership with CMC Markets Connect in 2024 brought
CFD trading capabilities into its ecosystem through an API arrangement. A
similar deal with GTN added bond trading for EEA customers the same
year.
Revolut X itself launched with ambitions targeting a $200 billion European crypto market, with Bashlykov cited
at the time as the driving force behind its expansion.
That API-first architecture is precisely what made the MCP
experiment possible. A broker whose trading infrastructure lives inside a
proprietary platform, rather than exposed through clean APIs, cannot simply
plug in Claude and tell it to start market-making.
A Different Kind of Product Roadmap
“The question is to which point it continues to make
sense bringing UI updates to the product if AI enabled flow allows so much more
capabilities,โ Bashlykov’s concluded.
Retail brokers have spent years competing on better charts,
faster buttons, and cleaner mobile apps. But a 30-minute session on Revolut
suggests that race may already be running in the wrong direction.
The firms that built their platforms for screens are now
watching rivals who built for APIs pull further ahead. And the gap is starting
to show.
Two engineers at Revolut’s crypto exchange built a fully
functional market-making system in about half an hour. Not a prototype. Not a
carefully prepared demo. Just a first idea, executed using Anthropic’s Claude
AI connected to the Revolut X trading API through a protocol called MCP.
The experiment is raising a question the retail trading
industry may not be ready to answer: if an AI agent can orchestrate an entire
trading workflow through a simple text prompt, what exactly is a broker’s
platform worth anymore?
Revolutโs Side Project That Broke the Product Roadmap
Nikita Ivanov and Vlad Kaminski, engineers on Revolut’s
crypto team, built the integration as a personal experiment, according to
Leonid Bashlykov, Revolut’s Head of Product for Crypto.
Leonid Bashlykov, Source: LinkedIn
“Two A-players on my team…were playing with Claude
automations and built the MCP as a side project, just for fun,” Bashlykov
wrote on LinkedIn. “Then we saw what it could actually do. And it broke
our product roadmap thinking.”
Bashlykov said he personally built a working market-making
strategy using Claude, MCP, and the Revolut X API in 30 minutes, with no
advance preparation. The workflow covered everything from portfolio screening
and position sizing to execution, monitoring and automated alerts. We are
talking about tasks that traditionally require separate tools, scripts, and a
fair amount of technical expertise to connect together.
MCP, or Model Context Protocol, is an open standard
developed by Anthropic that allows AI models to discover and use external tools
through a standardized interface. Rather than requiring custom-built
integrations for every API, MCP lets an AI agent connect to any compatible
system and reason across all of them simultaneously. Revolut X is currently
offering MCP access in beta.
Prompts Replace Dashboards
Linas Beliลซnas
The practical implication, according to Bashlykov, is that
complex trading strategies no longer require coding knowledge or platform
fluency. “Rebalance to 60/40 BTC/ETH if BTC dominance drops below
52%,” he wrote as an example. “That’s now a prompt.”
Fintech analyst Linas Beliลซnas, who covered the story in his
Weekly Fintech Pulse newsletter, framed it as a potential architectural shift.
“The agent reads portfolio data, checks conditions, pulls context – then
decides,” he wrote, contrasting MCP-connected agents with rule-based
trading bots that simply execute predefined instructions.
Beliลซnas noted that the workflow Bashlykov described,
handling inventory management, dynamic quoting, position sizing, execution,
monitoring and Telegram alerts, “normally requires a quant, a developer,
backtesting infrastructure, and weeks of iteration.”
Most AI trading tools currently on the market, both
Bashlykov and Beliลซnas agree, are still limited to price alerts, simple
queries, and basic automations.
Revolut engineers built an AI trading workflow with Claude in 30 minutes.
Now theyโre questioning whether they need to build product features at all ๐ค
Inside Revolut X, two engineers – Nikita Ivanov & Vlad Kaminski – started experimenting with Claude automations.
Just a sideโฆ pic.twitter.com/2YI65JygRv
โ Linas Beliลซnas (@linasbeliunas) March 4, 2026
What makes MCP different, the company claims, is
composability. A single connected workflow can simultaneously execute trades,
read news, track on-chain data flows and update dashboards. Revolut X functions
as the execution layer; Claude functions as the interface.
The Brokers Already Moving (and Those Who Aren’t)
Revolut is not entirely alone. CFD
broker ATFX partnered with data firm KX late last year to deploy an AI-driven
MCP server for real-time trading data, and LSEG connected its market data
feeds directly to ChatGPT in December 2025, allowing institutional users to
pull live information on demand. A FinanceMagnates.com analysis from November
2025 noted that crypto-native firms were already moving into AI’s data layer
while asking whether traditional brokers would follow.
The answer, so far, is: slowly. Firms like XTB, IG Group and
Plus500 have spent years and significant capital building proprietary trading
interfaces. XTB entered 2026 projecting $186 million in income while
doubling down on platform growth and client acquisition, a roadmap built around
the assumption that the interface is the product.
That assumption is now being tested. A comparison of client metrics across IG, CMC, Plus500 and XTB published
last August showed that client numbers are growing across the board, but
revenue per user varies dramatically. It signals that platform stickiness, not
just user count, drives broker economics. If an AI agent can replicate a
platform’s functionality through a plain-language prompt, the stickiness
argument weakens considerably.
Revolut’s Head Start
Revolut’s position in this shift is not accidental. The
company has systematically built API-first infrastructure across asset classes.
A partnership with CMC Markets Connect in 2024 brought
CFD trading capabilities into its ecosystem through an API arrangement. A
similar deal with GTN added bond trading for EEA customers the same
year.
Revolut X itself launched with ambitions targeting a $200 billion European crypto market, with Bashlykov cited
at the time as the driving force behind its expansion.
That API-first architecture is precisely what made the MCP
experiment possible. A broker whose trading infrastructure lives inside a
proprietary platform, rather than exposed through clean APIs, cannot simply
plug in Claude and tell it to start market-making.
A Different Kind of Product Roadmap
“The question is to which point it continues to make
sense bringing UI updates to the product if AI enabled flow allows so much more
capabilities,โ Bashlykov’s concluded.
Retail brokers have spent years competing on better charts,
faster buttons, and cleaner mobile apps. But a 30-minute session on Revolut
suggests that race may already be running in the wrong direction.
The firms that built their platforms for screens are now
watching rivals who built for APIs pull further ahead. And the gap is starting
to show.
