Anthropic has introduced the Model Hardware Standard (MHS) for hardware, designed to help AI agents driven by large models safely and swiftly operate physical devices. This is seen as a crucial step for AI to transition from the digital world to the physical world. The MHS standard originated from Anthropic's collaboration with the Howard Hughes Medical Institute's Janelia Research Campus and is now available as a research preview to the first batch of scientific research laboratories and advanced manufacturers. This standard can significantly reduce hardware integration time from weeks or even months to just hours or even minutes, while enabling researchers and engineers to coordinate autonomous 24/7 experiments and workflows. Agents can autonomously reason about experimental steps, update parameters, and even recover from hardware failures without human intervention. The MHS standard is widely adaptable and can be applied to any device with a programmable interface, regardless of the type of foundational model. It addresses challenges such as device communication and interaction with agents through standardized drivers, enabling agents to understand and operate unfamiliar devices. MHS provides three mechanisms for controlling hardware: MCP, command-line interface, and code files. In tests, the Claude agent demonstrated exploratory behavior similar to that of a scientist. Currently, several companies have built-in MHS support in their related devices. However, the MHS standard still has limitations, such as its inability to be compatible with hardware that lacks a programmable interface. Anthropic is continuously refining the standard and inviting early adopters to participate in related testing and integration efforts.