Open-Source Project Enables AI Agents to Use Simulation Models as Tools
The Virtual Sea Trial project, coordinated by Novia, focuses on the virtual commissioning of cruise ships using process simulations and AI-enhanced interfaces. Recently, researchers in the project bridged the gap between Artificial Intelligence (AI) and simulation by combining two industry standards: the Functional Mock-up Interface (FMI), which provides a framework for exchanging simulation models across different platforms, and the Model Context Protocol (MCP), which standardizes how applications provide context to AI agents and the tools they can use.
“Current developments in AI have been explosive, and we can expect many more innovations in the coming year. Keeping up with the pace of change is practically a full-time job. That is why we decided to make this an open-source project so anyone can contribute,” says Project Manager Mikael Manngård.
Running simulations with AI assistance
Traditionally, running simulations requires specialized software and expertise, limiting the use of simulation models to specialized individuals or teams within organizations. By giving AI agents access to simulation models and allowing users to interact with them through chat interfaces, virtually anyone can now set up and perform simulations. This can have an immediate positive impact on how information is shared within teams and across organizations.
The user interacts with the AI agent through an application, such as a chat interface like Claude Desktop. The AI agent can plan actions to answer the user’s questions, including requesting permission to run tools such as simulations in the background.
Two Standards: Hiding Complexity, Enabling Access, and Maintaining Accuracy
The Model Context Protocol (MCP), developed by Anthropic AI, is a community-adopted open standard that enables AI assistants to connect to external tools and data sources. AI assistants know complex concepts but cannot access real-time data or run calculations. MCP gives them these abilities.
“The Model Context Protocol fills in knowledge and skill gaps for AI agents and gives them additional functionalities that allow them to answer questions, perform actions, or make decisions based on real data,” says Project Researcher Christoffer Björkskog.
The Functional Mock-up Interface (FMI) is a standard for packaging and sharing simulation models so they can run in different tools and environments. It allows models to be exchanged and reused across different simulation platforms without exporting and compiling them multiple times. The FMI standard is widely adopted in the process simulation industry and has become the standard way of sharing models between parties.
In a sense, FMI and MCP are solving similar problems – making complex technical capabilities usable without exposing the underlying technical details. FMI abstracts the complexity of simulation models so they can run across different platforms without users needing to understand the internal mathematics or implementation. MCP abstracts the complexity of tool integration, so AI agents can access diverse capabilities without needing custom code for each connection. When combined, these two standards create a powerful synergy: FMI-packaged simulations become instantly accessible to AI agents through MCP, transforming sophisticated engineering analysis from a specialist activity into something anyone can use through natural conversation while maintaining the full technical integrity of the underlying models.