Anthropic previews a standard for AI agents to drive lab hardware

Anthropic announced a research preview of the Model Hardware Standard on 27 August 2026, a shared specification of software drivers and interfaces through which AI agents operate physical laboratory and manufacturing devices.

The problem it addresses is fragmentation. Each instrument ships its own proprietary interface with no common way to talk to it, so wiring several machines together has meant weeks or months of manual integration per setup. Anthropic says the standard reduces that to hours or minutes, citing an eight-hour integration at Carnegie Mellon against several weeks for a typical vendor setup. The standard is model-agnostic: any agent harness can reach a compliant device using standard protocols, including the Model Context Protocol.

Named research partners are Genentech, the Baker and Pinglay labs at the University of Washington, Carnegie Mellon University, HHMI Janelia Research Campus, QuEra Computing and Tetsuwan Scientific. Hardware vendors supporting it include Amazon Web Services, Automata, Danaher, Doosan Robotics, MBF Bioscience, QIAGEN, Tecan, Universal Robots, Hugging Face and Raspberry Pi. Supported device classes cover liquid handlers, robotic arms, plate readers, microscopes, lasers, cameras, thermocyclers and centrifuges.

It is not open source yet. Anthropic says it plans to open the standard after developing safety evaluations with this initial group, which places the decision about who may drive laboratory equipment with an agent inside a private preview for now.