Anthropic Brings Claude and Other AI Models Into Physical Equipment

Anthropic is moving its artificial intelligence work beyond screens and software with a new framework for connecting language models to physical equipment. The company released the Model Hardware Standard, or MHS, as a research preview on Thursday, August 27, 2026.

MHS connects advanced large language models, including Anthropic’s Claude, with physical objects. Anthropic says companies can use the framework to integrate AI into equipment in “hours or minutes,” a process that typically takes “weeks, if not months.”

A common language for equipment and AI

The framework allows multiple devices to connect with one another and communicate through commands such as “read.” Any hardware device can understand these commands and act on them, giving language models a way to work with equipment through a shared system.

MHS also enables “autonomous, round-the-clock experiments and workflows.” That feature gives labs and other organizations a way to connect AI systems with equipment that can carry out tasks without requiring people to guide every step.

Anthropic technical staff member Alek Kemeny compared the foundation of MHS to a familiar piece of technology infrastructure. “The MCP is ‘kind of like the USB for AI to software connection’,” Kemeny said.

MHS is built on the Model Context Protocol, or MCP, a universal, open standard for connecting data sources that Anthropic debuted in 2024. The hardware standard extends that approach from data and software connections to equipment that can receive commands and respond to them.

Designed to work across different models

MHS is model-agnostic, so it does not depend on Claude. It works with any large language model, including models built by other companies such as OpenAI and open-source models.

That detail matters for companies and researchers choosing how to build their systems. MHS gives equipment a shared connection method while allowing the language model behind the system to come from different developers.

Jonah Cool, Anthropic’s head of partnerships and deployment of science, said the company wants to keep scientists from being tied to one hardware or AI provider. “We want to avoid vendor lock-in for scientists,” Cool said.

Kemeny described a future in which the standard becomes part of the equipment itself. “In the future, scientists can buy these devices and out of the box it works. That’s just the process of adopting a standard,” he said.

Manufacturers still need to add the right interface

MHS does not connect to every existing piece of equipment automatically. Not all devices have a programming interface, which means they cannot receive or act on the commands required by the framework without changes.

Anthropic is working with device manufacturers to build new products with the necessary interface and pre-loaded MHS connections. The company is also helping manufacturers add MHS connections to products that already exist.

This work places hardware makers at the center of MHS adoption. The standard can provide a common way for devices to communicate, but each product still needs the interface that lets it understand commands and carry them out.

Early partners span science and robotics

Anthropic developed MHS in partnership with the HHMI Janelia Research Campus. Early access went to a “handful” of laboratories and hardware manufacturers working in areas such as biotech, robotics, and quantum computing.

The early partners include Genentech, Carnegie Mellon University, QuEra, Universal Robots, Amazon Web Services, Doosan Robotics, Danaher, and Hugging Face.

Those partnerships place MHS across several types of equipment and research. Biotech labs, robotics groups, and quantum computing organizations can all test how language models interact with physical systems through the same framework.

Anthropic’s research preview marks its first foray into physical AI. Instead of treating a language model as a tool that only generates text or software, MHS connects it to equipment that can read information, communicate with other devices, and act on commands.

The standard is still in its research preview, and existing equipment may need new interfaces before it can use MHS. But Anthropic’s plan is clear: make AI-connected hardware easier to adopt, give scientists more choice among language models, and create a shared connection between models and machines.

Stay connected via Google News

Follow us for the latest AI News