Anthropic Tests Claude With Robots And Labs
Anthropic has introduced a new framework that could let Claude AI agents operate robots and scientific equipment. The company calls it the Model Hardware Standard, or MHS.
The initiative extends AI agents beyond software and into physical environments. Moreover, MHS can connect Claude with microscopes, liquid handlers, robotic arms, and other programmable devices. It can also coordinate several instruments at once.
Anthropic developed MHS with HHMI Janelia Research Campus. Meanwhile, early partners include organizations across biotechnology, robotics, quantum computing, and manufacturing.
New Standard Connects Claude With Hardware
MHS addresses a major problem in laboratory automation. Traditionally, each instrument uses its own software interface. Therefore, connecting multiple machines often requires custom engineering and specialist support.
Anthropic says those integrations can take weeks or months. However, MHS aims to reduce that work to hours or minutes. The standard uses drivers that translate between computer systems and physical devices.
Furthermore, the system gives AI agents structured information about each device. That information can include capabilities, operating conditions, and safety limits. Agents can then discover equipment and coordinate operations through a common interface.
MHS also works with different AI models. As a result, the framework does not lock hardware developers into Claude. The system can use standard protocols such as Anthropic’s Model Context Protocol.
AI Moves Toward Autonomous Experiments
Anthropic has tested MHS across several scientific applications. For example, QuEra used the framework to give an AI agent control over part of a quantum computer’s laser system. The agent recovered the laser lock successfully 99.3% of the time without human intervention.
Meanwhile, Anthropic’s drug discovery work connected AI with liquid handlers, robotic arms, plate readers, and monitoring cameras. The company says MHS helped run certain experiments roughly three times faster.
However, Anthropic acknowledges that the technology still requires expert oversight. Claude can struggle with physical problems that software descriptions cannot fully capture. Therefore, researchers must intervene when equipment behaves unexpectedly.
The company plans to continue safety evaluations before making MHS open source. In addition, hardware makers are already exploring support for the standard. Consequently, Anthropic is positioning MHS as a potential foundation for AI-driven laboratories and industrial automation.
