Anthropic has previewed a standard that gives AI agents a single interface for controlling scientific instruments, from microscopes to lasers.
The Model Hardware Standard, or MHS, was developed with biomedical research institute HHMI. It replaces the disparate APIs used by lab machines – which force researchers to learn each device’s quirks and often do not interoperate – with one set of configuration commands.
HHMI has already built an agent that manages microscopes automatically, and quantum computing startup QuEra is using MHS with Claude to coordinate the lasers that program its machines. Anthropic says agents can also fix experiment errors and write instrument control scripts.
The standard includes safety guardrails, such as limits on laser strength, and will ship under an open-source license. AWS, Hugging Face and several industrial robot suppliers are working to integrate MHS.
Anthropic frames MHS as a step toward agents controlling manufacturing equipment, not just lab gear, putting the startup at the center of the push to make AI operate physical machines.
