Anthropic unveils new framework allowing AI agents to operate physical devices

Anthropic has unveiled a new standard called the Model Hardware Standard (MHS) that simplifies the way AI agents control machines, such as microscopes and quantum computers. This standard, developed in collaboration with the Howard Hughes Medical Institute (HHMI), allows AI agents to manage scientific instruments more efficiently.

The MHS replaces disparate device-specific APIs with a single, unified interface that provides standardized configuration commands. This simplification enables AI agents to coordinate scientific instruments more effectively, reducing the need for researchers to write custom code for each machine. Anthropic has demonstrated the MHS by building an AI agent that automatically manages microscopes and by partnering with startup QuEra Computing Inc. to coordinate lasers with its quantum computers.

The standard also makes it easier to automatically fix experiment errors and write instrument control scripts. MHS enables researchers to define the key properties of each instrument, such as the weight of a robotic arm or the wavelengths a sensor can detect, which helps AI agents identify the best way to use each machine and avoid configuration mistakes.

Safety guardrails can be set, like limiting the strength of laser beams to prevent damage to samples. Anthropic plans to release MHS under an open-source license, allowing its use not only in scientific instruments but also in manufacturing equipment. Major companies like Amazon Web Services, Hugging Face, and several industrial robot suppliers are already working to integrate MHS into their products.