Anthropic on Thursday opened a research preview for the Model Hardware Standard (MHS), a shared specification designed to let AI agents operate physical equipment in laboratories and manufacturing settings.
The standard enables AI agents to manage multiple instruments simultaneously, including microscopes, liquid handlers and robotic arms. Anthropic estimates MHS can reduce hardware integration time from weeks or months to hours or minutes. The specification is model-agnostic and compatible with any device that has a programmable interface.
MHS allows agent harnesses to access equipment via standard protocols such as the Model Context Protocol. It uses a standardized driver that translates between a computer's operating system and hardware devices through simple commands, making each device discoverable in a standard format without the need for custom integration programs.
Development of MHS was conducted in collaboration with the HHMI Janelia Research Campus. Anthropic has secured early support from several companies and platforms. Amazon Web Services will support MHS through its Strands Robots library, which connects AI agents to physical devices. Automata will add MHS support to LINQ, its lab automation platform, while Tecan will integrate MHS with its Fluent liquid handling platforms.
Universal Robots plans to support MHS on its robotics platform. Danaher is exploring how MHS could scale biomedical research using smart instruments and autonomous laboratories. Doosan Robotics is testing MHS with robotic arms for automated quality assurance, and QIAGEN is experimenting with MHS on its QIAsymphony Connect nucleic acid purification platform.
During the research preview, Anthropic will work with launch partners to build safety evaluations and develop protections for AI use with physical equipment. The company intends to open-source the standard after the preview period and will publish findings as part of its safe deployment guidance. Stakeholders across industries are being invited to join the research preview waitlist.
