What is Mcp Fmi

MCP-FMI stands for Model Context Protocol - Functional Mock-up Interface. It allows simulation models to be utilized as tools for LLM-based agents, enhancing their capabilities.

Use cases

Use cases for MCP-FMI include querying available simulation models, retrieving information about model inputs and outputs, simulating models with specific input changes, and integrating simulation capabilities into applications.

How to use

To use MCP-FMI, ensure you have Python 3.11 or higher and the uv package manager installed. Configure the ‘claude_desktop_config.json’ file to integrate the MCP-FMI server with Claude Desktop, specifying the command and arguments for your FMU directory.

Key features

Key features of MCP-FMI include integration of FMU simulation models, support for various queries related to simulation models, and the ability to simulate models with generated inputs.

Where to use

MCP-FMI can be used in fields such as simulation modeling, artificial intelligence, and systems engineering, particularly where LLM-based agents are employed.

Model Context Protocol - Functional Mock-up Interface
Makes your simulation models available as tools for LLM-based agents.

Novia UAS | Research Project | Contact

MCP-FMI

This is an unofficial MCP-integration of the FMPy package.

Table of Content

1. Overview
   1.1 Features
   1.2 Implemented Tools
   1.3 Standards
2. Prerequisites
3. Claude Desktop Integration
4. For Developers
5. Example Usage
6. Future Work
7. Citation
8. Acknowledgements
9. License

Overview

This package integrates FMU simulation models as tools for LLM-based agents through the MCP. By allowing simulation models to be accessed and orchestrated directly from chat interfaces, simulation models become accessible to virtually anyone. This means simulation capabilities can be shared across teams and organizations, including with non-experts, bringing the benefits of simulation to everyone you collaborate with.

Features

• 💬 Manage simulations from chat interfaces.
• 🛠️ Use simulation models as tools for LLM-based agents.
• 📈 Generate input signals for simulations from natural language.
• 📊 Visualize simulation results in browser.

Implemented tools

• fmu_information_tool retrieves information about the available FMU models.
• simulate_toolsimulates a single FMU model with default prameters and input signals. Returns the simulated outputs.
• simulate_with_input_tool simulates a single FMU model with the specified input signals. Returns the simulated outputs.
• create_signal_tool generates an input-sequence object for a single input.
• merge_signals_tool merges multiple signel objects that can be used as an input for a simulation.
• show_results_in_browser_tool visualizes simulation results in browser.

Standards

This package bridges two standards:

The Model Context Protocol (MCP) is an open protocol that standardizes how applications can provide context to Large Language Models (LLMs). MCP helps when integrating data and tools to LLM-based agents.

The Functional Mockup Interface (FMI) is a free standard that defines a container and interface to exchange dynamic simulation models across simulation platforms. A Functional Mockup Unit (FMU) is a file containing a simulation model that adheres to the FMI standard.

Prerequisites

• Python 3.11 or higher
• uv package manager
• Claude Desktop (for desktop integration)

Usage with Claude Desktop

Update the claude_desktop_config.json file with:

For developers

Intructions how to run the package in editable mode: CONTRIBUTING.md.

Example usage

Example natural-language queries:

• What simulation models do you have available?
• Please provide information about the model and its signals.
• Who created the model, and when was it last updated?
• Make a step change in the input input_name at 60s. Keep the other inputs constant with their default values.
• Simulate model_name with the generated inputs.

Short demonstration in Claude Desktop:

Future work

List of features to be implemented:

• Better support for Windows/Linux/MacOS FMU models.
• co_simulate_tool co-simulates multiple FMU models.
• Integration for OpenWebUI.
• show_results_as_artifact_tool return simulation results as interractive artifacts.
• Features for exporting results.

Citation

If you use this package in your research, please cite it using the following BibTeX entry:

@misc{MCP-FMI,
  author = {Mikael Manngård, Christoffer Björkskog},
  title = {MCP-FMI: MCP Server for the Functional Mock-Up Interface},
  year = {2025},
  howpublished = {\url{https://github.com/Novia-RDI-Seafaring/mcp-fmi}},
}

Acknowledgements

This work was done in the Business Finland funded project Virtual Sea Trial

License

This package is licensed under the MIT License license. See the LICENSE file for more details.