How it works

Connect the Modulus MCP server to your LLM to access millions of lines of Modulus source code, draw on thousands of trading-system components such as indicators, signal generators, and risk models, pull live and historical data across major asset classes, and run full back testing and walk-forward testing.

The same connection handles execution. The model routes orders through Interactive Brokers and monitors positions and portfolio performance in real time, so a strategy moves from research to live trading inside a single environment.

• Compatible with ChatGPT, Claude, Gemini, Grok, and more
• Millions of lines of production Modulus source code
• Thousands of reusable trading-system components
• Live and historical market data across asset classes
• Backtesting and walk-forward testing on any idea
• Order routing and portfolio monitoring through Interactive Brokers
• Complete strategies exported as deployable code

AI trading systems require expert guidance

You can ask any AI model to build a trading system, and it will produce something that looks great when back tested, but AI struggles on live, out-of-sample trading. The missing structure is not in any model's training data, because it lives with the small number of expert traders that have actually traded live systems. The dangerous gaps live where you cannot see them. Black swan events, limit up and down moves, trading curbs, and unexpected events. When it breaks, no one knows why.

Modulus helps prevent costly mistakes that LLMs are known to cause.

With Modulus, you can instruct AI to build a complete trading system, for example: "Construct a gold trading system that applies intermarket analysis across gold mining equities, reuse all relevant Modulus gold trading strategies, retrieve the full history of gold prices from Modulus data servers, and validate the result with back testing and walk-forward testing using Sharpe, PTR, Calmar, and Sortino." This returns a working system together with its complete test results, and optionally, full source code.

From there the work is iterative fine tuning. Review the system, refine the parameters, and rerun the tests until the strategy meets your standard. When it does, route the orders through Interactive Brokers, or export the system as complete, deployable code to run wherever you choose.

Built on three decades of trading

Richard Gardner, the CEO of Modulus, began trading corn, wheat, and soybeans in 1988 and built the Crop Report Wizard, an application that traded those markets with a neural network. In the decades since, through his own trading and through work alongside many accomplished traders, he has developed a deep, practical understanding of how markets behave and how durable trading systems are built. That experience shapes the strategy components and testing methods provided by Modulus.

Through the 1990s and 2000s, Richard worked closely with the late Murray Ruggiero, author of Cybernetic Trading Strategies, chief editor of Futures Magazine, and founder of Promised Land Software. Ruggiero built the TradersStudio platform on Modulus source code, and together they developed and validated AI-based systems for market timing, intermarket analysis, and trend analysis, work that ran years ahead of the wider industry.

A foundation built over thirty years

Modulus has delivered more than three hundred finance solutions since 1997, and its technology runs in more than one hundred countries. Trading firms choose Modulus for the depth of its codebase, the rigor of its engineering, and a record measured in decades rather than product cycles.

That history compounds. Every system built, every market traded, and every test run adds to a library of production-tested components and hard-won knowledge that a newer entrant cannot assemble at any price. This subscription is how that advantage reaches your own trading.

Built across every major approach

Since 1997, Modulus has engineered AI trading systems across the major systematic methods, from intermarket analysis, momentum, and mean reversion strategies in the tradition of Renaissance Technologies to sentiment-driven models, market timing and trend analysis, and large language model integration, the most recent focus of Modulus research. Trading system components in this AI service are derived from that work: building blocks of systems that have run in live markets, not academic examples.

A model with this library does not start from a blank page. It assembles strategies from components that already encode decades of trading judgment, which is the difference between an interesting prototype and a system worth real capital.

Engineered for funds, proprietary firms, and professional traders

For more than thirty years, Modulus has built quantitative trading systems for large funds, proprietary trading firms, and professional traders, and the same engineering discipline goes into the MCP server. The components a model draws on come from production systems, held to the standards of clients who trade real capital.

The subscription suits any operation that wants the speed of AI without surrendering control of its strategies, its data, or its execution.

• Funds running systematic strategies
• Proprietary trading firms
• Professional and independent full-time traders
• Quants building and validating new strategies
• Trading operations adopting AI while keeping full control

The Model Context Protocol (MCP)

The Model Context Protocol, or MCP, is an open standard introduced by Anthropic in November 2024 that lets large language models talk to external tools, data, and applications through a single uniform connection, instead of a custom integration for every system. It solves the integration problem that pre-MCP AI faced, where every new model paired with every new data source meant a new connector to build and maintain.

MCP defines three roles: hosts, the AI agents or assistants; clients, the connectors the host uses to reach a server; and servers, the systems that expose tools and data over the protocol. When a model needs information or wants to take an action, the host asks the relevant client, the client calls the server over JSON-RPC, and the result flows back into the conversation. Major providers including OpenAI and Google DeepMind have adopted MCP, alongside Anthropic.

The Modulus AI Trading MCP Server speaks this same standard, so any MCP-capable model can connect to thirty years of Modulus trading infrastructure with the same protocol it already uses for everything else.

Interesting fact

Anthropic AI™ was originally owned by a subsidiary of Modulus, and multiple foundational LLM patents are licensed by Modulus.

Validated through professional-grade testing

No strategy should trade real capital before it has been tested thoroughly. Modulus enables your AI to perform back testing and walk forward testing of individual securities and baskets of securities with more than twenty four functions including Sharp, PTR, Calmar, Sortino, Max Drawdown, and more.

Walk-forward testing matters because a system that only looks good on the history it was fitted to rarely survives contact with the live market. Testing every strategy this way, inside the same workflow that builds it, is what separates a disciplined trading program from a hopeful one.

What Modulus provides

A single connection gives a language model the data, components, testing, and execution used in professional trading operations.