Anthropic Releases Standard Linking Claude to Robots and Lab Equipment

Anthropic has announced a research preview of the Model Hardware Standard (MHS), a specification that lets AI models, including Claude, autonomously control physical lab and manufacturing hardware. It's the company's first product aimed at the physical world rather than software alone.

What MHS Is

The Model Hardware Standard is a shared specification that makes lab and manufacturing equipment, such as microscopes, robotic arms, lasers, and liquid dispensers, recognizable and controllable by AI agents without writing dedicated software for each manufacturer separately. The standard relies on simple read and write commands that any device can interpret, while information about safety parameters and control methods is stored as reference files generated from natural-language descriptions.

Agents access hardware through the Model Context Protocol, a command-line interface, and an API, allowing a single command to orchestrate multiple devices at once. Anthropic stresses that the solution is model-agnostic - MHS can be used not only by Claude but also by other companies' models, including OpenAI's, as well as open-source models.

How It Works in Practice

At biotech company Genentech, researchers connected Claude to a liquid dispenser, a robotic arm, and a plate reader to automate a BCA protein assay. The model independently selected optimal flow parameters, including 140 microliters per second for water and 10 microliters per second for higher-viscosity samples, without input from a lab automation specialist.

At the University of Washington, deploying remote device monitoring and AI-supervised qPCR processes took less than a week, compared to integration projects that usually take months. At the HHMI Janelia research center, seven separate manufacturer programs for two-photon microscopy equipment were unified into a single interface, cutting the time to connect a new device from a multi-day project to a few minutes.

Limitations and Safety

Anthropic acknowledges that Claude still struggles with physical, chemical, and biological intuition, particularly when troubleshooting unusual equipment failures that require real lab experience. Before releasing the standard as open source, the company is running safety tests with partners and preparing a separate roadmap to limit the risk of errors when controlling physical devices.

Claude interacts with experiments and hardware in an exploratory way, much like a scientist would - Anthropic, Model Hardware Standard description

We want to avoid locking scientists into a single hardware vendor - Jonah Cool, head of scientific partnerships and deployment, Anthropic

What It Means for Labs and Companies

For research labs and manufacturing plants, this could mean moving away from costly, closed integrations built separately for each equipment maker. If MHS is indeed released openly, smaller research teams would gain access to automation that, until now, only large pharmaceutical and technology corporations could afford.

Barclays analysts cited alongside the announcement estimate that the market for AI-powered robots could reach a value of one trillion dollars by 2035, with labs and factories among the first practical applications of the technology beyond chat and text generation.

For now, access to the MHS research preview is limited to an initial group of scientific labs and industrial manufacturers on a waitlist, with full release of the standard depending on the results of ongoing safety testing.