FAQ
Straight answers about the spec, who is building support, and the fact that this catalog is a community project — not Anthropic.
Is this directory run by Anthropic?
No. This is an independent community project. Anthropic does not operate, endorse, or support this site. Apply for the official research preview at modelhardwarestandard.com.
What is the Model Hardware Standard?
The Model Hardware Standard (MHS) is Anthropic’s research-preview specification, announced 27 August 2026 with HHMI Janelia, for AI agents to discover and safely operate programmable physical devices. A standard driver uses read/write primitives, network discovery, and natural-language safety tags. Agents reach devices over MCP, a CLI, or generated code files. Access is apply-only; open source is planned after safety work.
Is MHS open source?
No. As of 28 August 2026 MHS is a gated research preview. Anthropic says it will open-source the standard after safety evaluations with launch partners. There is no official public GitHub org for the spec yet.
How does MHS relate to MCP?
MHS and MCP are complementary, not competing. MHS describes how AI agents discover and safely control physical hardware (cameras, robots, lab equipment) through a standard driver format. MCP (Model Context Protocol) is a software interface that exposes tools and data sources to agents. An agent often uses MCP to call an MHS driver, making them work together: MCP exposes tools, MHS describes hardware. Neither replaces the other.
What's the difference between MCP and MHS?
MCP is a protocol for connecting AI models to software tools and data sources. MHS is a specification for connecting AI agents to physical hardware devices. Think of it this way: MCP helps agents talk to your database or API; MHS helps agents control your robot arm or microscope. An MHS driver can be exposed through MCP, but MHS also supports CLI and code file interfaces.
Which companies are building MHS support?
Named in Anthropic’s announcement: AWS (Strands Robots), Automata (LINQ), Danaher, Doosan Robotics, MBF Bioscience (ScanImage), QIAGEN (QIAsymphony Connect), Tecan (Fluent), Universal Robots, Hugging Face (LeRobot), and Raspberry Pi (Camera MHS Driver). Waters Corporation joined the MHS research program on 13 October 2026 (LC-MS instruments; Waters AI Link MCP for waters_connect is in early access). Labs include Genentech, HHMI Janelia, Carnegie Mellon, University of Washington Baker and Pinglay labs, QuEra, and Tetsuwan Scientific.
What devices does MHS cover?
Any device with a programmable interface. Official examples include cameras, robot arms, microscopes, centrifuges, pipette robots, spectrometers, and incubators. Early work also includes liquid handlers, plate readers, qPCR, and quantum lasers. Hardware without a programming interface is out of scope until vendors add drivers.
How do I apply for the MHS research preview?
Apply on Anthropic’s official site at modelhardwarestandard.com, not through this directory. This site indexes public names and links; it does not proxy access to the spec.
Is there a public MHS GitHub repository?
Not for the spec itself. Hugging Face LeRobot is the main public repo to watch for MHS-shaped code; support is announced, not confirmed merged. Unrelated LCD-show “MHS35” Raspberry Pi displays are not Model Hardware Standard.
Where is the machine-readable MHS catalog?
A public catalog feed is at gofastmhs.com/api/catalog for software. This website is the human index.
Subscribe for updates
Spec drops, new drivers, catalog changes. No ads. We email when something actually shipped.