MoziChem-Hub is a modular MCP server hub that exposes core chemical engineering and chemistry tools as MCP servers. It leverages the FastMCP framework to package individual modules (eos-models, flash calculations, and thermodynamic properties) as MCP endpoints that can be consumed by AI agents, IDEs, or client applications via standardized MCP APIs. The README demonstrates how to run each module using UV (Python virtual environment management) for isolated, dependency-resolved execution. To integrate with client tooling, you configure multiple MCP servers in a single configuration, enabling seamless orchestration of EOS modeling, flash calculations, and thermodynamic property lookups through a unified hub. Clients such as Claude Desktop or VS Code can connect to these servers by referencing the appropriate server names and execution commands. The available modules provide: eos-models-mcp for advanced thermodynamic properties and EOS root analyses, flash-calculations-mcp for bubble/dew point and general flash computations, and thermodynamic-properties-mcp for component property lookups from multiple data sources.

Prerequisites:

• Python 3.8+ installed on the host machine
• Git installed to clone the repository (optional if you already have the code)
• Internet access to install Python dependencies via the project configuration
• Knowledge of absolute paths for your project directory

Installation steps:

1. Obtain the MoziChem-Hub code (clone or download the repository): git clone https://github.com/your-org/mozichem-hub.git cd mozichem-hub

2. Create and activate a Python environment (recommended): python -m venv venv source venv/bin/activate # Linux/macOS venv\Scripts\activate.bat # Windows

3. Install dependencies listed in the project (the repository typically includes a pyproject.toml or requirements.txt). If a pyproject.toml is present, install via: pip install -e .

   or install development dependencies as documented in the repo

4. Prepare your configuration. Use absolute paths when configuring MCP servers. For example, create a configuration file (config.json) containing the mcpServers block shown in the mcp_config section. Adjust the directory paths to your local setup.

5. Run the MCP servers using the provided commands (as shown in the README examples). For each module, launch via UV in your environment: uv --directory /absolute/path/to/mozichem-hub run eos-models-mcp.py uv --directory /absolute/path/to/mozichem-hub run flash-calculations-mcp.py uv --directory /absolute/path/to/mozichem-hub run thermodynamic-properties-mcp.py

6. Verify that the servers are reachable via their MCP endpoints and begin accepting requests from MCP clients.

Notes:

• If you prefer direct Python execution instead of UV, you can run the Python script directly and set PYTHONPATH accordingly. See the README examples for the direct Python configuration.
• Always use absolute paths in your configuration to avoid path-resolution issues across environments.

Tips and common issues:

• Ensure your Python environment has all dependencies required by each module (eos-models, flash-calculations, thermodynamic-properties).
• When using UV, ensure your project directory contains the pyproject.toml to enable automatic environment management.
• If you modify module scripts or add new MCP modules, update the mcpServers section accordingly with the correct script paths and server names.
• For Claude or VS Code integrations, provide the correct absolute path to the Mozichem-Hub project and ensure the command is set to uv with the proper --directory flag.
• If you encounter environment variable issues, you can declare env blocks per server similar to the Python example in the README, e.g., "env": {"PYTHONPATH": "/absolute/path/to/mozichem-hub"}.
• Monitor logs for each server to diagnose startup errors or missing dependencies; typical problems include missing Python packages or mismatched Python versions.