mozichem

MoziChem-MCP provides specialized MCP servers for chemical engineering calculations. The repository hosts two independent servers: eos-models-mcp, which offers Equation of State modeling (Peng-Robinson, Soave-Redlich-Kwong, Redlich-Kwong, van der Waals), fugacity calculations, and thermodynamic property predictions; and flash-calculations-mcp, which performs vapor-liquid equilibrium, multi-component phase equilibrium, and related flash calculations. These servers are designed to be queried by AI assistants via the Model Context Protocol, enabling sophisticated thermodynamic and phase behavior queries to be answered programmatically. To use them, run each server with the uvx runtime and point your MCP-enabled assistant to the appropriate command/arguments, for example using the uvx command with the mozichem-mcp package name and the specific server slug. Once running, you can instruct your AI assistant to perform tasks like fugacity calculations, phase equilibrium analysis, or flash calculations for given mixtures and conditions, and the assistant will format requests using MCP to obtain results from the respective server.

Prerequisites:

• Python 3.13 or higher
• uv package manager (recommended)

Install from source:

# Clone the repository
git clone https://github.com/sinagilassi/mozichem-mcp.git
cd mozichem-mcp

# Install using uv (recommended)
uv sync

# Or install using pip (editable install)
pip install -e .

Install from PyPI (when available):

pip install mozichem-mcp

Usage after installation:

• Run the EOS Models MCP server:

# Using uvx with the published package
uvx --from mozichem-mcp mozichem-mcp-eos-models

# Or run directly with Python (if installed locally)
python -m mozichem_mcp.mcp.eos_models

• Run the Flash Calculations MCP server:

# Using uvx with the published package
uvx --from mozichem-mcp mozichem-mcp-flash-calculation

# Or run directly with Python (if installed locally)
python -m mozichem_mcp.mcp.flash_calculation

If you prefer not to install locally, you can also use the package via uvx with the --from mozichem-mcp and the respective server slug, as shown above.

Tips and notes:

• Ensure your Python environment uses Python 3.13+ as required.
• The uvx runtime is recommended for running these MCP servers; it manages package contexts for MCP integration.
• The two servers are independent; you can run them separately and configure your MCP/AI tool to query either one depending on the calculation type.
• When integrating with AI assistants, reference the server slugs mozichem-eos and mozichem-flash in your configuration, as shown in the example Claude Desktop configuration in the README.
• If you encounter environment issues, verify that the Mozich/MCP package is installed in the active environment and that the uvx context is correctly resolving mozichem-mcp-eos-models and mozichem-mcp-flash-calculation modules.
• For advanced usage, you can invoke the servers directly with Python modules, which is helpful for local development or debugging.