Anthropic's New Standard Lets AI Control Physical Hardware

Anthropic built the Model Context Protocol to let AI models talk to software tools without custom code for every connection. Now the company wants the same fix for physical hardware. On Thursday, Anthropic detailed the Model Hardware Standard, or MHS, a spec designed to let AI agents read data from and control equipment like microscopes and robotic arms in research labs and factories.

The problem it targets is familiar to anyone who has run a lab. A microscope from one manufacturer, a robotic arm from another, a spectrometer from a third: each with its own API, its own data format, its own control software. Wiring them together for an AI system to use has traditionally taken weeks or months of custom engineering. Anthropic built MHS with standardized drivers so devices show up in a common format, which the company says can cut that integration time down to hours or minutes.

The spec was developed with HHMI Janelia Research Campus and is going out first as a research preview. Anthropic has already run tests with lab partners, including on quantum computing hardware, to see how AI agents handle real physical equipment rather than just software interfaces.