Harnessing AtomisticSkills for Agentic Atomistic Research
9 affiliations
- 1Department of Materials Science and Engineering , Massachusetts Institute of Technology , Cambridge , MA 02139 , USA
- 2Department of Chemical Engineering , Massachusetts Institute of Technology , Cambridge , MA 02139 , USA
- 3Department of Chemistry , Kookmin University , Seoul 02707 , Republic of Korea
- 4Harvard University , Department of Chemistry and Chemical Biology , Cambridge , MA , 02138 , USA
- 5Department of Chemistry , Massachusetts Institute of Technology , Cambridge , MA 02139 , USA
- 6Department of Nuclear Science and Engineering , Massachusetts Institute of Technology , Cambridge , MA 02139 , USA
- 7MACE Base Server
- 8Shell Information Technology International Inc ., Texas 77082 , United States
- 9Shell International Exploration & Production Inc ., Texas 77079 , United States
Abstract
Computational materials science and chemistry span vast knowledge domains and fractured software ecosystems. Although large language models (LLMs) have demonstrated research capabilities, scaling monolithic agents to manage the rigor and complexity of atomistic research remains a challenge. Here, we introduce AtomisticSkills, an open-source harness framework that empowers general-purpose AI coding agents to conduct atomistic research across materials science, chemistry, and drug discovery. By hierarchically decomposing scientific workflows into agent skills and tools, AtomisticSkills provides agents with modular, extensible, and plug-and-play research capabilities. The framework integrates more than 100 human-curated multidisciplinary skills, including database access, thermodynamics and kinetics modeling, and diverse simulation engines employing machine learning interatomic potentials (MLIPs) and density functional theory (DFT). We validate its functional coverage against scientific literature and demonstrate robust orchestration capabilities across diverse scientific campaigns: generative design of Li-ion solid-state electrolytes, high-throughput screening of metal-organic frameworks for CO2 capture, autonomous MLIP benchmarking and fine-tuning, multi-stage structure-based virtual screening for drug design, multimodal X-ray diffraction pattern analysis, and screening of Fe-oxide catalysts for oxygen evolution reaction. AtomisticSkills provides a critical agent infrastructure towards building fully autonomous AI scientists.
Related
- RetroAgent: Harnessing LLMs to Search Over Structured Memory for Agentic Retrosynthesis Planning
- El Agente Potente: High-Throughput Agentic Atomistic Simulations
- ETALON: Verified Execution and Persistent Research Memory for Agentic Computer-Aided Drug Design
- A traceable agentic framework for chemistry research query and analysis with ThermoML database
- Molecular augmented dynamics: Generating experimentally consistent atomistic structures by design
BibTeX
@misc{deng2026harnessing,
title = {{Harnessing AtomisticSkills for Agentic Atomistic Research}},
author = {Bowen Deng and Bohan Li and Matthew Cox and Hoje Chun and Juno Nam and Artur Lyssenko and Sathya Edamadaka and Jurğis Ruža and Xiaochen Du and Nofit Segal and Jesus Diaz Sanchez and Mingrou Xie and Ty Perez and Yu Yao and Miguel Steiner and Sauradeep Majumdar and Charles B. Musgrave and Anirban Chandra and Abhirup Patra and Detlef Hohl and Connor W. Coley and Ju Li and Rafael Gómez‐Bombarelli},
howpublished = {arXiv (Cornell University)},
year = {2026},
month = may,
doi = {10.48550/arxiv.2605.24002},
url = {https://arxiv.org/abs/2605.24002},
note = {CC BY-SA},
}
RIS
TY - UNPB
TI - Harnessing AtomisticSkills for Agentic Atomistic Research
AU - Bowen Deng
AU - Bohan Li
AU - Matthew Cox
AU - Hoje Chun
AU - Juno Nam
AU - Artur Lyssenko
AU - Sathya Edamadaka
AU - Jurğis Ruža
AU - Xiaochen Du
AU - Nofit Segal
AU - Jesus Diaz Sanchez
AU - Mingrou Xie
AU - Ty Perez
AU - Yu Yao
AU - Miguel Steiner
AU - Sauradeep Majumdar
AU - Charles B. Musgrave
AU - Anirban Chandra
AU - Abhirup Patra
AU - Detlef Hohl
AU - Connor W. Coley
AU - Ju Li
AU - Rafael Gómez‐Bombarelli
PY - 2026
DA - 2026/05/18
PB - arXiv (Cornell University)
DO - 10.48550/arxiv.2605.24002
UR - https://arxiv.org/abs/2605.24002
AB - Computational materials science and chemistry span vast knowledge domains and fractured software ecosystems. Although large language models (LLMs) have demonstrated research capabilities, scaling monolithic agents to manage the rigor and complexity of atomistic research remains a challenge. Here, we introduce AtomisticSkills, an open-source harness framework that empowers general-purpose AI coding agents to conduct atomistic research across materials science, chemistry, and drug discovery. By hierarchically decomposing scientific workflows into agent skills and tools, AtomisticSkills provides agents with modular, extensible, and plug-and-play research capabilities. The framework integrates more than 100 human-curated multidisciplinary skills, including database access, thermodynamics and kinetics modeling, and diverse simulation engines employing machine learning interatomic potentials (MLIPs) and density functional theory (DFT). We validate its functional coverage against scientific literature and demonstrate robust orchestration capabilities across diverse scientific campaigns: generative design of Li-ion solid-state electrolytes, high-throughput screening of metal-organic frameworks for CO2 capture, autonomous MLIP benchmarking and fine-tuning, multi-stage structure-based virtual screening for drug design, multimodal X-ray diffraction pattern analysis, and screening of Fe-oxide catalysts for oxygen evolution reaction. AtomisticSkills provides a critical agent infrastructure towards building fully autonomous AI scientists.
SN - 2331-8422
ER -