Autonomous Universal Research Assistant (AURA): Agentic AI meets nanoHUB's FAIR Workflows and Data

Agentic artificial intelligence is increasingly impacting scientific research, from ideation and planning to execution. However, most research assistants target specific domains and rely on predetermined toolsets, limiting scalability and community contribution. We introduce AURA, a multi-agent assistant that integrates a large language model with nanoHUB's expert-contributed, FAIR simulation tools and data. AURA discovers and orchestrates these sources of information by reasoning over their metadata at runtime. Thus, new tools and data become available as they are published. With case studies in protein folding, high-pressure materials properties, and melting temperature calculations, we demonstrate capabilities for scientific planning, tool discovery, workflow composition, and autonomous execution. Learning from prior simulation results, AURA optimizes tool inputs and reproduces a recently published result in minutes, a 20x–30x speedup over expert researchers. AURA illustrates a scalable approach to autonomous research assistants that encourages community contributions and will result in an organic growth in capabilities.