Anthropic on August 27 released a research preview of the Model Hardware Standard (MHS), a specification that lets AI agents read data from and control physical laboratory and manufacturing equipment such as microscopes, robotic arms, liquid handlers, and lasers5,11,14.
MHS introduces standardized drivers that make each device discoverable in a common format, allowing devices and agents to communicate across networks without requiring bespoke integration software2,6. Anthropic says the standard works with any device that has a programmable interface and is model-agnostic10. The company says MHS cuts hardware integration time from weeks or months down to hours or minutes.
The effort began as a collaboration between Anthropic and the HHMI Janelia Research Campus in Ashburn, Virginia3. Anthropic Technical Staffer Alek Kemeny said the project was inspired by observing neuroscientist Arco Bast coordinate rotating laser beams, microscopes, cameras, and other components through a common interface during a memory-formation experiment.
MHS extends Anthropic's Model Context Protocol (MCP), an open standard launched in November 2024 for two-way communication between large language models and external software tools, into the domain of physical hardware12. Agents can control hardware through MCP, a command-line interface, or code files via API. Agents can also coordinate multiple devices and save workflows as deterministic scripts that run without a language model in the loop.
Anthropic presented several partner demonstrations. At Genentech, Claude coordinated a liquid handler, a robotic arm, and a plate reader to automate a protein assay, optimizing pipetting parameters for different liquids on its own. At Carnegie Mellon University, researchers connected a liquid handler, plate reader, robotic arm, and monitoring cameras across three computers with incompatible interfaces; the team said the drivers and orchestration layer took about eight hours to build, compared to the several weeks a vendor setup would typically require. At QuEra, a quantum computing company using neutral atoms, Claude used MHS across hundreds of automated runs to develop a control program that recovers a laser's lock after disruptions, succeeding in 695 out of 700 attempts (99.3 percent) in a blind test, running entirely without the language model. These examples come from partner projects presented by Anthropic and have not been independently verified.
The research preview is open to a first group of scientific research labs and advanced manufacturers. Anthropic said it is sharing an early version with partners to build safety evaluations before an eventual open-source release. The company is also developing a physical safety roadmap to guard against misuse13.
The partner ecosystem already taking shape includes Amazon Web Services, which will support MHS through a private pre-release of its Strands Robots package during the preview period. Doosan Robotics is testing MHS with its robotic arms for automated quality assurance and multi-robot coordination. QIAGEN is running a proof of concept on its QIAsymphony Connect nucleic acid purification platform. Tecan is adding MHS support to its Fluent liquid handling platform, and Universal Robots has received early access with plans to integrate MHS into its robotics platform. Hugging Face is adding MHS support to its robotics library LeRobot, and Raspberry Pi is enabling MHS integration across multiple products after successful tests with its Camera MHS driver. Automata, Danaher, MBF Bioscience, and Tetsuwan are also building or exploring MHS integrations.
Anthropic acknowledged that Claude's spatial and physical reasoning still has limits requiring expert oversight. During the preview, the company plans to build additional safety evaluations with partners and will release findings as part of deployment guidance when it open-sources the standard.
ANALYSIS The breadth of the initial partner list, spanning robotics vendors, life-science instrument makers, a quantum computing firm, and open-source hardware, positions MHS as a horizontal integration layer rather than a vertical solution for a single industry.