Development and Evaluation of an Agentic Suite of Tasks Covering a Computational Chemistry Curriculum

Computational chemistry, increasingly enabled by high-performance computing, accelerates research across many areas of chemistry. However, heterogeneous software packages, input formats, and output conventions create a steep learning curve that also complicates simulation scripting and automation. Large language models (LLMs) can lower this barrier by helping users prepare calculations, execute workflows, and interpret results. We present ASSAY (Agentic Simulation Suite for Automated Chemistry), an open-source agentic toolkit for quantum chemistry designed for undergraduate CC education. To align ASSAY with current curricula, we analyzed syllabi from 17 courses at public and private universities in and outside the United States. ASSAY supports example-driven learning through input generation and output interpretation, while its local and remote workflows can introduce students to distributed computing, batch queues, and file management. In a benchmark spanning 12 skills and 8 LLMs, ASSAY achieved a mean pass rate of 94.04%, with six models achieving at least 98%. These results establish ASSAY as a reproducible, integrity-first foundation for making computational chemistry more accessible to students and reliable for researchers.