Fujitsu Social Digital Twin

You can run Fujitsu's Social Digital Twin MCP Server to expose the Digital Rehearsal API to MCP-compatible LLM clients. The server communicates over standard input/output, letting you start simulations, retrieve results, and analyze data using natural language options provided through MCP tools.

How to use

To interact with the Fujitsu Social Digital Twin MCP Server, launch it in a standard I/O MCP client workflow. The server exposes a set of tools that let you list simulations, start a simulation, fetch results, analyze traffic, compare scenarios, and generate configurations from natural language. You can operate it from a command line interface or through an MCP-compatible LLM client that communicates via standard input/output.

How to install

Prerequisites: you need Python 3.13 or higher and access to Fujitsu API Gateway with an API key. You also need an MCP-compatible LLM client (for example, Claude Desktop). Install and run steps are shown below.

# 1) Install the runtime environment (example using uv as the Python runtime helper)
# Create a virtual environment and activate it
python -m venv .venv
source .venv/bin/activate  # Unix/macOS
.venv/Scripts/activate     # Windows

# 2) Install dependencies
pip install -r requirements.txt

# 3) Set environment variables (adjust values for your setup)
export FUJITSU_API_BASE_URL=https://apigateway.research.global.fujitsu.com/sdtp
export FUJITSU_API_KEY=your_api_key_here

# 4) Run the MCP server
python -m fujitsu_sdt_mcp

If you prefer a streamlined setup script, you can use the provided setup script to initialize the environment for you.

chmod +x setup.sh
./setup.sh

Additional configuration and usage notes

Claude Desktop and other MCP-compatible clients can connect to the server using the standard I/O protocol. When configuring Claude Desktop, specify the runtime as a command that launches the MCP server via Python, and pass the required environment variables.

// Example configuration for Claude Desktop
{
  "command": "python",
  "args": ["-m", "fujitsu_sdt_mcp"],
  "env": {
    "FUJITSU_API_BASE_URL": "https://apigateway.research.global.fujitsu.com/sdtp",
    "FUJITSU_API_KEY": "your_api_key_here"
  }
}

Resources and tools

The server provides a set of tools to manage and analyze simulations, including listing simulations, starting simulations, retrieving results, obtaining metrics, accessing simulation data, analyzing traffic, comparing scenarios, and generating configurations from natural language.

Available tools include: list_simulations, start_simulation, get_simulation_result, get_metrics, list_simdata, get_simdata, analyze_traffic_simulation, compare_scenarios, create_natural_language_simulation_config.

Example MCP client workflows

You can perform common tasks with the following tool invocations to drive simulations and analysis.

tool list_simulations

tool start_simulation {"simdataId": "your_simdata_id"}

tool analyze_traffic_simulation {"simulationId": "your_simulation_id", "region": "Tokyo", "timeRange": "morning rush hour", "scenario": "traffic optimization"}

tool create_natural_language_simulation_config {"description": "I want to run a traffic optimization simulation during morning rush hour in Tokyo"}

Available tools

list_simulations

Retrieve a list of available simulations from the Digital Rehearsal MCP server.

start_simulation

Start a selected simulation using a given simdataId.

get_simulation_result

Fetch results for a completed or in-progress simulation.

get_metrics

Retrieve key simulation metrics and performance indicators.

list_simdata

List available simulation data entries for selection.

get_simdata

Retrieve specific simulation data by ID.

analyze_traffic_simulation

Analyze traffic-oriented simulations by region, time, and scenario.

compare_scenarios

Compare outputs across different simulation scenarios.

create_natural_language_simulation_config

Generate simulation configurations from natural language input.