jcliugroup/ai-computational-chemist

AI Computational Chemist, or AICC, is a collection of instruction files that teach an AI agent how to do computational chemistry and materials science work. It is not one program you run, it is a set of skills, written as documents an agent can read and follow, that work with tools like Claude Code, Codex, Cursor, or opencode. The main use it is built for is helping researchers respond to peer review. When a scientific paper's reviewers ask for extra calculations, an AI agent using these skills can read the manuscript and the reviewer comments, figure out which comments actually need new calculations, run those calculations using the same scientific methods the paper already used, check whether each result really answers the reviewer's concern, and then help draft the written response and supporting material. A human is expected to approve the plan before calculations start and approve the final draft before anything is sent out. If a new calculation contradicts what the original paper claimed, the process stops and flags it to the researchers instead of hiding or reframing the result. The skills also work on their own, outside the peer review use case, for general computational chemistry tasks like density functional theory, quantum chemistry, molecular dynamics, machine learning potentials, phonon calculations, and reaction kinetics modeling. For bigger projects, there is an additional layer called the research orchestrator, which tracks a whole project as a set of linked tasks with dependencies, records what has been decided and why, and keeps different tasks from stepping on each other when multiple agents are working at once. The repository is organized into three kinds of building blocks: procedures that guide multi-step work, tools that operate specific chemistry software packages like VASP, Gaussian, and LAMMPS, and a knowledge library of chemistry background material the other two can draw on.