Using MLatom with AI agents

MLatom can be driven by AI agents that understand atomistic simulations – they can install MLatom, choose and set up methods, run calculations, and assemble workflows from a natural-language request. There are several options, depending on whether you want a ready-to-use online assistant, an agentic system installed locally, or skills for your own AI coding assistant.

Protomia

Protomia is the main AI assistant for atomistic simulations built on MLatom. It runs online, with nothing to install, on the Aitomistic Hub and Aitomistic Lab@XMU (the latter for users with an academic email), and carries out complete simulations from a natural-language request.

MLatom Skills

MLatom Skills is an open, agent-agnostic collection of skills – short, tested recipes that teach any AI coding assistant (such as Claude, Cursor, and GitHub Copilot, as well as Protomia) how to perform a specific MLatom task: installing MLatom, running a method, or reproducing a published workflow. It is a new and growing collection, and open to contributions – if you have an MLatom workflow worth sharing, you are welcome to add it. Point your assistant at the relevant SKILL.md (start with the install skill if you are setting MLatom up).

The MLatom repository also includes an AGENTS.md file that orients an AI agent working in a cloned copy: what MLatom is, how to install it, and where to find the skills and documentation.

Aitomia

Aitomia is an agentic system for AI-driven atomistic and quantum-chemical simulations that academic users can install and run locally.