Anthropic is now extending its expertise in software interoperability into the physical world. The AI startup has unveiled a new software specification called the Model Hardware Standard (MHS), designed to enable AI assistants like Claude to collaborate more efficiently with robots, scientific instruments, and manufacturing equipment, marking a significant step in the expansion of large language models into physical environments.
According to a Thursday statement reported by Bloomberg, Anthropic has released MHS as a "research preview," opening it up to developers in the scientific, robotics, and manufacturing sectors for testing. This allows them to validate how Claude interacts with the physical world and establish necessary safety mechanisms before a full rollout. Industry participants can currently join a waitlist to request access, and Anthropic has indicated that the framework will be open-sourced after the preview period, though no specific timeline has been disclosed.
The launch of MHS comes at a time when the commercial value of integrating AI with the physical world is drawing widespread attention. Barclays previously published a report projecting that the AI-driven robotics and autonomous machine market could grow into a trillion-dollar sector by 2035. Google, OpenAI, and Nvidia are also concurrently developing AI models and software tailored for robotics applications.
The Core Logic of MHS: Enabling AI to "Understand" Hardware
At its essence, MHS provides AI assistants with a standardized information storage mechanism for understanding how hardware operates, replacing paper manuals or implicit knowledge held only by a few experts. For example, with a factory robotic arm, hardware manufacturers can explicitly define safety boundaries for AI control within the MHS specification, such as limiting movement speed or angles, thereby granting AI operational capabilities while mitigating safety risks. Anthropic engineer Alek Kemeny stated: "What MCP did for the software world, MHS will achieve for the hardware world. It is especially valuable in scientific laboratory settings, where dozens or even hundreds of devices are often running simultaneously."
Benchmarking Against MCP: From Software Connectivity to Hardware Connectivity
The strategic logic behind MHS is directly aligned with Anthropic's previously launched Model Context Protocol (MCP). MCP has already become an industry standard for AI assistants interacting with third-party software applications, enabling Claude and competitor ChatGPT to access mainstream apps like Gmail, Google Calendar, and Slack. With MHS positioned as the "hardware version of MCP," Anthropic aims to replicate MCP's success in physical device connectivity, building a hardware ecosystem moat around Claude through open standards and pushing the standard toward industry-wide adoption after open-sourcing.
Additionally, Anthropic has revealed that MHS has already completed early testing with several industry organizations, including partners such as Amazon Web Services (AWS), Danaher, Hugging Face, and Raspberry Pi. In a promotional video released by Anthropic, a scientist from biotechnology company Genentech sent a self-designed experimental PDF document to Claude, which then autonomously activated hardware equipped with the MHS specification to execute the entire experimental workflow.
Jonah Cool, Head of Partnerships and Deployment for Anthropic's Life Sciences division, commented: "Many times, scientific research cannot progress because researchers do not know how to use the equipment, or the operation is too complex. MHS empowers Claude to become an enabling tool for scientists, helping them use the right equipment in a professional manner."