harnessing
in
your agentic IDEs
A composable framework for AI-driven atomistic research. Chain modular skills across MLIPs, DFT, drug discovery, and generative AI — in any coding copilot.
Strictly typed, battle-tested operations exposed via MCP servers. Fixed signatures for relax, MD, literature search, DFT I/O, and more.
Mid-level tutorials combining tool calls to solve focused research problems. Each ships with SKILL.md, helper scripts, and runnable examples.
High-level, publication-scope research campaigns chaining multiple skills. From computational materials science workflows to building atomistic ML models.
Click any skill to view its full documentation and example runs.
Pre-built pipelines for complex multi-stage research objectives.
Each environment is fully isolated to prevent dependency conflicts. Environments marked MCP expose tools directly to your AI agent; others are used via skills scripts.
MCP Server Environments
| MCP Server | Tools | Conda Env | Core Libraries | Purpose |
|---|---|---|---|---|
| base | base-agent | pymatgen, ase, rdkit, packmol | MP queries, VASP I/O, structure tools | |
| mace | mace-agent | mace-torch | MACE models (MP, OMAT, MATPES, MH) | |
| matgl | matgl-agent | matgl, dgl | CHGNet, M3GNet, TensorNet | |
| fairchem | fairchem-agent | fairchem-core | UMA, ESEN models (OMAT, OMol, OC20) | |
| atomate2 | atomate2-agent | atomate2, jobflow-remote | Remote DFT workflows via NERSC/HPC | |
| smol | smol-agent | smol | Cluster expansion + Monte Carlo | |
| drugdisc | drugdisc-agent | rdkit, autodock-vina, meeko | Docking, ADMET, fingerprints | |
| mattergen | mattergen-agent | MatterGen, PyG, lightning | Generative crystal design | |
| adit | adit-agent | ADiT, lightning, hydra, PyG | All-atom diffusion generation | |
| diffcsp | diffcsp-agent | DiffCSP++, hydra, PyG, pyxtal | Symmetry-constrained crystal generation |
Script-Only Environments (no MCP server — invoked via conda run)
| Conda Env | Core Libraries | Purpose |
|---|---|---|
| xrd-agent | DARA, pymatgen | XRD phase analysis & Rietveld refinement |
| orca-agent | scine_utilities, scine_readuct, ase | DFT/CC single-points, TS optimization (ORCA) |
| calphad-agent | pycalphad, pymatgen | CALPHAD phase & property diagrams |
| phasefield-agent | fipy, scipy, imageio | Phase-field simulations (Allen-Cahn, Cahn-Hilliard) |
| nmr-agent | nmrsim, nmrglue | NMR prediction & Wasserstein deconvolution |
| react-ot-agent | PyTorch | Transition state generation via React-OT |
In-depth guides for developing new skills, managing environments, and data visualization.
AtomisticSkills is designed to work directly with agentic IDEs. Follow these simple steps:
git clone git@github.com:learningmatter-mit/AtomisticSkills.git && cd AtomisticSkills