Crystalyse

A provenance-enforced scientific agent for computational materials design. Crystalyse orchestrates computational tools through natural language while ensuring every numerical result traces to explicit calculations, addressing the core challenge of AI hallucination in scientific domains.

From a terminal prompt like "Suggest a new Na-ion battery cathode", the system autonomously computes capacity (193 mAh/g) and voltage (3.7 V) in ~90 seconds—despite having no pre-coded battery workflows. The agent reasons about which fundamental calculations to chain together, then derives electrochemical properties.  The system orchestrates established tools—SMACT for compositional screening, Chemeleon for structure generation, MACE foundation models for energy calculations, PyMatGen for stability analysis, while enforcing that every numerical property must trace to explicit tool invocations, with audit trails showing which calculation produced each result.

The goal is both to leverage LLM creativity and accelerate computational aspects of materials design, so scientists can focus their effort on the challenging discovery questions. This is early work and a proof-of-concept framework that makes advanced computational methods feel as natural as starting with a written prompt.

December 1st, 2025 — v1.0.0 stable release and preprint now available

Read the preprint •  View the code •  See updates  •  contact 

Built by Ryan Nduma, Hyunsoo Park, and Aron Walsh at Imperial College London.