Automate AMS Workflows with Python

• Start with a simple AMS geometry optimization in Python
• Build AMS input blocks with PLAMS Settings
• Convert AMS text input or run files to PLAMS format
• Run many AMS jobs in parallel
• Summarize job results with JobAnalysis
• Use AMS as an ASE calculator
• Run molecular dynamics from Python
• Extract frames and data from AMS trajectories
• Generate, rescore, and filter conformers
• Train ReaxFF, DFTB, and ML models with ParAMS workflows
• Automate Reactions Discovery with ReaxFF

How PLAMS works

1. Build the system: create or import molecules, surfaces, slabs, periodic systems, or trajectories.
2. Define the calculation: use PLAMS Settings to choose the AMS task, engine, properties, and workflow options.
3. Run jobs: execute calculations locally, in parallel, or through job schedulers such as SLURM and TORQUE.
4. Read results: extract energies, structures, spectra, trajectories, properties, and custom data from AMS result files.
5. Automate the workflow: loop over structures, engines, settings, molecules, conformers, reaction paths, or training data.

Python workflows in AMS2026

AMS2026 expands the Python Examples library with reusable scripts for AMS engines, molecular dynamics, reaction paths, spectroscopy, materials modeling, COSMO-RS, ParAMS, active learning, and external engines. Most AMS users can run these scripts directly with amspython.

For users who want a separate Python project, the Amsterdam Modeling PySuite page explains how to install AMS Python packages with uv or pip.

Related AMS tools

Use AMSinput to set up jobs graphically and export them as PLAMS scripts when you want to customize or scale the workflow. Use ParAMS for parameter fitting, COSMO-RS with pyCRS for thermodynamics scripting, and ASE calculators when you want AMS to interact with the wider Python materials ecosystem.