Superpower Daily
Chinese Models Take More Than 60% of OpenRouter Tokens, Driven...Model-Serving Platforms Reach 6.1% of AI-Spending Businesses, W...Gemini Omni 1.1 Flash Adds 40-Second Scene Extensions and 4K Fi...Databricks Says 300M Chart-JSON Pipeline Tops Four Multimodal B...London Surgeons Remove 11mm Brain Tumour With Live AI Anatomy M...Nvidia’s $63.1 Billion Receivables Add a Cash-Flow Caveat to It...Chinese Models Take More Than 60% of OpenRouter Tokens, Driven...Model-Serving Platforms Reach 6.1% of AI-Spending Businesses, W...Gemini Omni 1.1 Flash Adds 40-Second Scene Extensions and 4K Fi...Databricks Says 300M Chart-JSON Pipeline Tops Four Multimodal B...London Surgeons Remove 11mm Brain Tumour With Live AI Anatomy M...Nvidia’s $63.1 Billion Receivables Add a Cash-Flow Caveat to It...
Toolspublished

Anthropic Puts Model-Agnostic MHS in Preview for AI-Controlled Machinery

Anthropic’s proposed interface could reduce the bespoke work of connecting agents to equipment, but broad adoption still rests on future safety work and an open-source release.

By 2 min read
Anthropic Puts Model-Agnostic MHS in Preview for AI-Controlled Machinery
Sign in to save and react

Listen to this story

The audio brief

About 1:30
0:001:30
Read transcript
Anthropic is testing a new interface that lets AI agents communicate with and control programmable physical equipment, from laboratory instruments to factory machines and robots. It’s called the Model Hardware Standard, or MHS, and it’s currently available only as a controlled research preview for a select group of science, robotics, and manufacturing organizations. The goal is to replace some of the bespoke integration work required to connect an AI system to each piece of hardware. MHS is designed to be model-agnostic, meaning an equipment operator would not have to use Anthropic’s own Claude models. The company compares the idea to USB-C: a common connection standard, rather than a machine or an AI model itself. The proposed scope is broad. It includes microscopes and liquid-handling equipment in laboratories, quantum-computing hardware, manufacturing machinery, and robot arms. But physical control creates a more serious safety boundary than a typical software integration. MHS includes specifications that can restrict an agent from using particular hardware, and Anthropic says it will work with trusted partners on safety before making the interface generally available. Anthropic also plans to open-source MHS in the future. That would extend its Model Context Protocol approach from software integrations to physical equipment, and give manufacturers a path to adopt the interface across industries. The key thing to watch is whether Anthropic completes that safety work and follows through on the open-source release; for now, MHS remains a limited preview, not an industry standard.

Story brief

3 key points

Anthropic is testing Model Hardware Standard (MHS), a model-agnostic interface for AI agents to operate programmable lab, manufacturing, quantum-computing, and robotic equipment. The August 27 announcement is limited to a select set of science, robotics, and manufacturing organizations, with safety controls for restricting hardware actions. MHS could reduce bespoke integrations and let device makers support multiple...

  1. 01

    MHS is currently a controlled research preview, not a generally available standard for device manufacturers.

  2. 02

    Target equipment includes microscopes, liquid handlers, quantum-computing hardware, factory machines, and robot arms.

  3. 03

    The interface is intended to work beyond Anthropic’s Claude models, potentially reducing provider lock-in for hardware operators.

Laboratory instruments, factory machinery and robots may become easier for AI agents to operate through a common interface. Anthropic has introduced the Model Hardware Standard, or MHS, for agents that communicate with and control programmable devices.

The immediate rollout is limited: MHS is available as a research preview to a select group of organizations in science, robotics and manufacturing. Anthropic announced the standard on August 27, positioning it as a way to reduce the time companies spend setting up and integrating hardware.

A common language for machines

MHS is designed for devices with programmable interfaces. Its intended scope includes machinery used in scientific research and advanced manufacturing, and Anthropic has described the interface as model agnostic, so organizations would not be limited to Claude models. The company compares the approach to USB-C: a standard for moving information between devices, rather than a machine or model itself.

Equipment MHS is meant to cover

  • Microscopes and liquid-handling equipment used in laboratory work.
  • Quantum-computing hardware, manufacturing machines and robot arms.

Physical access raises a safety boundary

The framework is also meant to give scientists and engineers a way to specify how an AI model should avoid using particular hardware, helping prevent mishaps. Anthropic plans to work with trusted partners on safety before making MHS generally available. That makes its current availability a controlled research phase, not a general-purpose deployment path for device makers.

An open-source path is planned

Anthropic says it plans to open-source MHS in the future, allowing device manufacturers in any industry to adopt it. The effort extends the company’s earlier Model Context Protocol, which defines rules for models interacting with software programs. MHS carries that standard-setting approach from software connections to physical equipment; the next material step is whether Anthropic completes the safety work and opens the standard to manufacturers.

Sources

  1. cnbc.comAnthropic pushes into physical world with new standard to help AI agents operate machines
  2. wired.comThis Is How Anthropic Thinks AI Agents Should Navigate the Physical World

Related coverage

Continue reading