It’s been three weeks since MetaQuotes joined the MCP craze and the platform provider has revealed that users have already processed over 1 trillion tokens through its built-in AI assistant. On the surface, this indicates swift adoption across a platform that commands a dominant share. Yet, processing vast quantities of text is quite different from generating genuine financial value.
The platform provider’s July release came with a built-in AI assistant for the MT5 client terminal.
Speaking to Finance Magnates, Christoforos Theodoulou, Chief Business Officer at MetaQuotes, has called the tool an “orchestrated coding agent powered by LLMs.” He explained it allowed traders and developers to “analyse code, plan multi-step actions, and automate routine tasks across the workflow.”
Because these capabilities sit directly within the primary interface and are offered for free, it has surely catalyzed adoption.
It is More than Just Tinkering
Since early 2026, several major players have released MCP servers to allow AI models to interact directly with trading platforms.
Interestingly, Claude appears in nine out of ten brokers that were investigated by Finance Magnates.
This trend has raised questions about whether it will fundamentally change the way traders interact with markets.
- Are Third-Party Bridge Providers Being Priced Out By MetaQuotes?
- MetaQuotes Launches MetaTrader.com as Standalone Financial Data Portal
- GMG Prime Taps MetaQuotes’ Ultency to Bring Institutional Liquidity to MT5
MetaQuotes claims that “thousands of MT5 users are applying the AI assistant to various tasks,” extending from chart analysis to automated strategy development. Set against MT5’s global user base of millions, a few thousand is a modest fraction, which puts the 1-trillion-token figure into context.
It shows that early adopters are not merely testing the tool once or twice. Instead, they are putting the system through heavy, intensive daily use.
Tokens Don’t Measure Success
However, token metrics measure computational throughput rather than anything practical. Tokens simply record the volume of text moving into and out of a model.
Someone can easily burn through a huge chunk of tokens by feeding poorly structured prompts into the system or repeatedly attempting to debug a single piece of code.
Similarly, a novice trader might generate extensive token counts asking basic questions about market trends without gaining any real analytical edge.
So, while the 1-trillion-token figure demonstrates that traders are willing to experiment, it fails to answer a fundamental question. Has this flurry of activity translated into faster development, fewer programming bugs or better trading results?
