Anthropic's Model Hardware Standard: AI Agents Get a Unified Driver for Lab & Factory Devices

Anthropic has opened a research preview of the Model Hardware Standard (MHS) — a shared specification that lets AI agents safely operate physical devices like microscopes, liquid handlers, and robotic arms. The goal is to cut the typical weeks-to-months hardware integration time in labs and factories down to hours or minutes.

How MHS works

MHS introduces a standardized driver that translates between a device and an AI agent. The driver exposes a simple set of primitives — commands like read (e.g., "get temperature") and write (e.g., "set temperature") — that any programmable device can understand. It also makes devices discoverable across networks without custom "translator" programs.

Critical for safe operation, the driver can include natural-language tags describing device characteristics that aren't obvious from code (like the weight of a robot arm). Users can add these tags manually or by chatting with an agent that interviews them about the setup. From these tags, MHS automatically generates a reference file with device capabilities, adjustability, and enforced safety limits — essentially giving the agent the information that used to live in paper manuals.

Three control mechanisms

Once connected, agents can control hardware through three mechanisms:

• MCP (Model Context Protocol) — standard protocol for agent-to-tool communication.
• Command line interface — direct terminal control.
• Code files (APIs) — chaining driver commands into scripts for fast, long-running operations.

These mechanisms work together to orchestrate multiple devices via a single line of code. Agents can sequence steps across instruments, monitor results in real time, and adjust parameters as conditions change — even recovering from hardware errors without human intervention.

Anthropic reports that in testing, Claude interacted with experiments in an exploratory manner, similar to a scientist. One observed example: Claude adjusted a laser, then used a camera to visually assess the effect — a loop that demonstrates the agent's ability to reason about physical actions.

MHS is being shared with a first group of scientific research labs and advanced manufacturers, including a collaboration with HHMI Janelia Research Campus. Anthropic plans to build safety evaluations and best practices before making the standard open source.

Who should apply

The preview is open to scientific research labs, robotics teams, electronics manufacturers, and advanced manufacturing facilities. Applications are being accepted now via Anthropic's website.