Abstract

AI scientists are emerging computational systems that serve as collaborativepartners in discovery. These systems remain difficult to build because they arebespoke, tied to rigid workflows, and lack shared environments that unifytools, data, and analyses into a common ecosystem. In omics, unified ecosystemshave transformed research by enabling interoperability, reuse, andcommunity-driven development; AI scientists require comparable infrastructure.We present ToolUniverse, an ecosystem for building AI scientists from anylanguage or reasoning model, whether open or closed. TOOLUNIVERSE standardizeshow AI scientists identify and call tools, integrating more than 600 machinelearning models, datasets, APIs, and scientific packages for data analysis,knowledge retrieval, and experimental design. It automatically refines toolinterfaces for correct use by AI scientists, creates new tools from naturallanguage descriptions, iteratively optimizes tool specifications, and composestools into agentic workflows. In a case study of hypercholesterolemia,ToolUniverse was used to create an AI scientist to identify a potent analog ofa drug with favorable predicted properties. The open-source ToolUniverse isavailable at https://aiscientist.tools.