Previewing the Model Hardware Standard

Details

Anthropic is opening a research preview of the Model Hardware Standard (MHS), a standardized framework co-developed with HHMI Janelia Research Campus that lets AI agents operate physical lab and manufacturing equipment. The core problem MHS solves is device fragmentation: most lab instruments have proprietary interfaces, requiring specialists to build one-off integrations that can take weeks or months. MHS introduces a standardized driver layer using simple primitives ('read', 'write') that makes any device with a programmable interface discoverable and operable by any AI agent using standard protocols including the Model Context Protocol. In practice, a Claude agent can simultaneously orchestrate a microscope, liquid handler, and robotic arm in a single workflow — reasoning through steps, updating parameters in real time, and in some cases recovering from hardware errors without human intervention. Integration time drops from weeks to hours. The standard is model-agnostic and harness-agnostic: it works with any agent that can use MCP, not just Claude. Anthropic is sharing the early version with partners in science, robotics, electronics, and manufacturing to co-develop safety evaluations and best practices before open-sourcing the standard. Cited use cases include drug discovery experiments and laser calibration on quantum computers. Development began as a two-party collaboration between Anthropic and Janelia, lending the project scientific credibility beyond a pure product announcement.