---
title: "Anthropic releases Model Hardware Standard for lab agents"
slug: "anthropic-releases-model-hardware-standard-for-lab-agents"
published: "2026-09-01"
beat: "Tools"
tags: ["Tools", "Launches"]
creator: "Agentry Newsroom"
editor: "Susanne Sperling, Editor — Human in the Loop"
tools: ["Claude (Anthropic)", "Perplexity Sonar"]
creativeWorkStatus: "verified"
dateReviewed: "2026-09-01"
aiActArticle50: "compliant"
humanView: "https://agentry.news/launches/anthropic-releases-model-hardware-standard-for-lab-agents"
agentView: "https://agentry.news/agent/anthropic-releases-model-hardware-standard-for-lab-agents"
---# Anthropic releases Model Hardware Standard for lab agents

> Anthropic PBC unveiled a research preview of Model Hardware Standard (MHS) on August 27, 2026, a framework that enables AI agents to operate physical devices in scientific research and advanced manufa

*Drafted by an AI agent. Verified by Susanne Sperling, Editor — Human in the Loop. [AI policy](/ai-policy).*

Anthropic PBC released a research preview of Model Hardware Standard (MHS) on August 27, 2026, a framework designed to let Claude AI agents safely interact with physical devices in scientific research and advanced manufacturing environments [Reuters](https://www.reuters.com/technology/anthropic-unveils-new-framework-allowing-ai-agents-operate-physical-devices-2026-08-27/).

## What MHS Does

The standard provides developers with a way to test how Claude interacts with the physical world and build in necessary safeguards before deploying agents to control equipment [Bloomberg](https://www.bloomberg.com/news/articles/2026-08-27/anthropic-tests-new-way-for-claude-to-work-with-robots-and-scientific-lab-tools). Anthropic named microscopes and robotic arms as examples of devices the framework is designed to support [Reuters](https://www.reuters.com/technology/anthropic-unveils-new-framework-allowing-ai-agents-operate-physical-devices-2026-08-27/).

Anthropologic said it was "**kicking off the first phase of the research preview** for Model Hardware Standard (MHS)" and sharing "**an early version of the MHS with partners**" to help build safety evaluations before open sourcing [Reuters](https://www.reuters.com/technology/anthropic-unveils-new-framework-allowing-ai-agents-operate-physical-devices-2026-08-27/), [X/@AnthropicAI](https://x.com/AnthropicAI/status/2093038426140651791).

## Roadmap: Open Source After Safety Work

The company has committed to making MHS open source in a later phase, once Anthropic completes safety evaluations with early partner feedback. This staged rollout lets the lab isolate risks and failure modes while the standard is still in controlled hands rather than deployed broadly across research institutions and manufacturing floors.

## Why This Matters for the Agent Economy

MHS represents a concrete developer tool for a category of agents that can directly manipulate the real world—moving beyond text generation or code execution into physical intervention. Agents that operate laboratory equipment or manufacturing robotics are subject to equipment damage, safety hazards, and contamination risks. A standard that lets teams test agent behavior in these domains before deployment addresses a core constraint: **how do you verify an agent won't break a $500,000 microscope or misalign a robotic arm?**

The framework is developer-facing infrastructure, shipped as a research preview but already usable by Anthropic's partner pool. It sits at the intersection of agent product development and safety validation—the kind of infrastructure that unblocks real deployment of agentic systems in labs and manufacturing.

Anthropic's move follows a broader trend of AI labs releasing agent tools and frameworks designed to make agentic systems safer and more controllable in high-stakes domains. By publishing a standard rather than keeping the capability proprietary, Anthropic is attempting to shape how the broader developer community approaches hardware control.