Anthropic wants to do for physical hardware what its Model Context Protocol did for software
decoder ·
Anthropic has introduced the Model Hardware Standard (MHS), a unified interface that lets AI agents read from and control physical devices such as microscopes and robotic arms. Inspired by its Model Context Protocol for software, MHS uses standardized drivers to make devices discoverable in a common format, cutting integration time from weeks or months to hours or minutes. In partner tests at Genentech, Claude coordinated lab equipment to automate a protein assay and optimized parameters, but it struggled with physical cause and effect when bubbles formed, requiring human guidance to fix the process. At Carnegie Mellon, researchers connected incompatible devices across three computers in about eight hours, and Claude independently improved an experimental measurement curve on a second run. At QuEra, Claude developed a control program for stabilizing a laser that succeeded in 695 of 700 blind tests (99.3 percent) while running without a language model. Anthropic notes the results are unverified partner projects, and expert oversight remains necessary due to Claude's limited physical reasoning. The company plans safety evaluations, a physical safety roadmap, and an open-source release, with manufacturers including AWS, Doosan Robotics, and Universal Robots already testing or supporting MHS.
decoder