About OpenMM MCP Server
OpenMM MCP Server is an MCP server for molecular dynamics simulations, built on OpenMM with Abacus DFT integration. It accepts natural language requests to submit, manage and run simulation tasks, aimed at researchers in computational chemistry and biophysics.
Key features
- 01Creation and control of OpenMM simulations
- 02Abacus DFT quantum mechanical calculation support
- 03Preconfigured protein and membrane setup templates
- 04CUDA, OpenCL and CPU platform options
- 05Asynchronous task execution and status tracking
Use cases
- 01Run molecular dynamics simulations of water or proteins
- 02Set up membrane protein simulations with pressure control
- 03Check and analyze completed simulation tasks
MCP support
The server exposes tools for simulation creation, task control and result analysis to MCP clients, so an LLM can drive OpenMM and Abacus through natural language.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
We haven’t collected a package or remote endpoint for this server. Follow the instructions in the project README.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMEOpenMM MCP Server
🧬 OpenMM & Abacus MCP Server
A comprehensive Model Context Protocol (MCP) server for molecular dynamics simulations using OpenMM and DFT calculations with Abacus. This server provides a natural language interface for complex molecular simulations through LLM integration.
✨ Features
- Complete OpenMM Integration: Support for all OpenMM features including advanced integrators, barostats, and constraints
- DFT Calculations: Abacus DFT engine integration for quantum mechanical calculations
- Pre-configured Templates: Ready-to-use setups for protein simulations, membrane systems, and more
- Advanced Sampling: Metadynamics, free energy calculations, and enhanced sampling methods
- GPU Acceleration: CUDA and OpenCL platform support for high-performance computing
- Task Management: Asynchronous task execution with persistence and monitoring
- Natural Language Interface: Interact with complex simulations using simple English commands
🚀 Quick Start
Installation
# Clone the repository
git clone <repository_url>
cd openmm-mcp-server
# Install dependencies
pip install -r requirements.txt
# Optional: Install OpenMM for actual simulations
conda install -c conda-forge openmm
# Test installation
python test_mcp_server.py
Roo Code Integration
Add to your Code/Cline/Claude MCP settings:
{
"mcpServers": {
"openmm-server": {
"command": "python",
"args": ["run_openmm_server.py"],
"cwd": "path-to\\openmm-mcp-server",
"alwaysAllow": [
"create_md_simulation",
"create_advanced_md_simulation",
"setup_protein_simulation",
"setup_membrane_simulation",
"create_dft_calculation",
"control_simulation",
"get_task_status",
"list_all_tasks",
"analyze_results"
]
}
}
}
Important: Replace the path with your actual project path!
🛠️ Available Tools
Basic Tools
create_md_simulation- Create simple MD simulationscreate_dft_calculation- Create DFT calculationscontrol_simulation- Control simulation execution (start/stop/pause)get_task_status- Check task statuslist_all_tasks- List all tasksanalyze_results- Analyze simulation results
Advanced Tools
create_advanced_md_simulation- Full OpenMM feature support (50+ parameters)setup_protein_simulation- Pre-configured protein simulation templatessetup_membrane_simulation- Membrane protein simulation setup
💬 Usage Examples
Simple Water Simulation
"Run a molecular dynamics simulation of a water molecule at 300K for 10000 steps"
Advanced Protein Simulation
"Set up a protein production simulation for 100 nanoseconds at physiological temperature 310K, using Amber19 force field with TIP3P-FB water model, 0.15M salt concentration, and GPU acceleration"
Membrane Protein Simulation
"Create a POPC membrane protein simulation for 50 nanoseconds with surface tension control"
📊 OpenMM Parameters Reference
Integrators
LangevinMiddle- Langevin middle integrator (recommended)Verlet- Verlet integrator (NVE ensemble)Brownian- Brownian dynamicsVariableLangevin- Variable step size LangevinVariableVerlet- Variable step size VerletNoseHoover- Nose-Hoover thermostat
Force Fields
- Amber14:
["amber14-all.xml", "amber14/tip3pfb.xml"] - Amber19:
["amber19-all.xml", "amber19/tip3pfb.xml"] - CHARMM36:
["charmm36_2024.xml", "charmm36/water.xml"]
Water Models
tip3p- TIP3P water modeltip3pfb- TIP3P-FB water model (recommended)tip4pew- TIP4P-Ew water modelspce- SPC/E water model
Platforms
CUDA- NVIDIA GPU (fastest)OpenCL- General GPUCPU- CPU computationReference- Reference implementation
Nonbonded Methods
PME- Particle Mesh Ewald (recommended for periodic systems)NoCutoff- No cutoff (small systems)CutoffNonPeriodic- Non-periodic cutoffCutoffPeriodic- Periodic cutoffEwald- Traditional Ewald summation
Constraints
None- No constraintsHBonds- Hydrogen bond constraints (recommended)AllBonds- All bond constraintsHAngles- Hydrogen angle constraints
Barostats
MonteCarloBarostat- Isotropic pressure controlMonteCarloAnisotropicBarostat- Anisotropic pressure controlMonteCarloMembraneBarostat- Membrane system pressure control
Precision
mixed- Mixed precision (recommended)single- Single precisiondouble- Double precision
🔧 Configuration
Environment variables:
export TASK_DATA_DIR="./simulation_data"
export MAX_CONCURRENT_TASKS=4
export DEFAULT_OPENMM_PLATFORM="CUDA"
export LOG_LEVEL="INFO"
export LOG_FILE="./mcp_server.log"
📁 Project Structure
openmm-mcp-server/
├── run_openmm_server.py # Startup script
├── test_mcp_server.py # Test script
├── requirements.txt # Dependencies
├── README.md # English documentation
├── README_CN.md # Chinese documentation
├── src/
│ ├── server_new.py # Main server
│ ├── advanced_md_tools.py # Advanced MD tools
│ ├── task_manager.py # Task management
│ ├── openmm_engine.py # OpenMM engine
│ ├── abacus_engine.py # Abacus DFT engine
│ ├── config.py # Configuration management
│ ├── tools/ # MCP tool implementations
│ ├── resources/ # MCP resource implementations
│ └── utils/ # Utility functions
├── tests/ # Test files
├── examples/ # Example code
├── docs/ # Documentation
└── task_data/ # Task data directory (auto-created)
📚 Additional Documentation
- Installation Guide - Detailed installation instructions
- Usage Guide - Comprehensive usage examples
- Roo Code Integration - Detailed integration guide
- OpenMM Parameters - Complete parameter reference
- Project Analysis - Technical analysis and improvements
🔍 Troubleshooting
Common Issues
-
"OpenMM not found" warning
- This is normal, server will run in mock mode
- For actual simulations, install OpenMM:
conda install -c conda-forge openmm
-
"command not found" error
- Check if Python is in PATH
- Verify project path is correct
- Try using absolute paths
-
Permission errors
- Ensure read/write permissions for project directory
- Check if task_data directory is writable
Verify Configuration
# Test server startup
python run_openmm_server.py
# Test MCP functionality
python test_mcp_server.py
🤝 Contributing
- Fork the repository
- Create a feature branch
- Make your changes
- Add tests
- Submit a pull request
🆘 Support
If you encounter any issues:
- Run the test script:
python test_mcp_server.py - Review the documentation
- Check log files (if configured)
- Submit an issue on GitHub
📄 License
This project is licensed under the GNU General Public License v3.0.
Happy simulating! 🎉
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.