LLM‐Powered AI Agents in Materials Science: From Autonomous Design to Autonomous Experimentation

Abstract

The rapid advancement of artificial intelligence is reshaping materials research, moving it beyond trial‑and‑error approaches toward more intelligent and autonomous discovery. AI agents powered by large language models (LLMs) have emerged as a promising component of this transition. Unlike conventional predictive models, they can integrate reasoning, knowledge retrieval, planning, and tool use to coordinate multiple stages of materials research, thereby strengthening the connection between conceptual design and experimental validation. In this review, recent advances in material‐science agents from the complementary perspectives of autonomous design and autonomous experimentation are highlighted. The technical foundations are first discussed, with emphasis placed on modular agent architectures, LLM‐centered reasoning and planning, memory, tool use, and domain‐adaptation strategies that connect general‐purpose foundation models with specialized materials knowledge. Representative applications are then systematically reviewed across information extraction, hypothesis generation, inverse design, laboratory automation, and closed‐loop optimization. Although these agents demonstrate the capacity to integrate multimodal information and accelerate iterative design‑experiment cycles, several challenges remain, including limited platform interoperability, insufficient availability and quality of materials data, and the lack of rigorous benchmarking and evaluation protocols. Finally, a staged roadmap is proposed to facilitate the development of a standardized, reliable, and increasingly autonomous ecosystem for next‑generation materials research.