Interoperable Natural Language Interfaces for Self-Driving Labs via Model Context Protocol

Wenyu Zhang, Maria Politi, Rama El-khawaldeh, Jason Hein

Keywords: Self-Driving Labs; Large Language Models; Model Context Protocol; Interoperable Robotic Control

TL;DR: An MCP server enabling natural language control across Python-based self-driving labs

Abstract: The development of self-driving laboratories (SDLs) is accelerating materials discovery by automating synthesis and characterization with robotic platforms and diverse instrumentation. However, the lack of standardized software interfaces hinders their broader adoption and interoperability. A parallel challenge arises with building agentic AI that can reliably control diverse physical systems. In this work, we present an architecture that integrates the Model Context Protocol (MCP)—an open standard for tool interaction with large language models (LLMs)—with an interoperable laboratory orchestration layer. This enables natural language interaction across the full spectrum of SDL functionality, from direct instrument control to closed-loop workflow design. We demonstrate its capabilities through two representative use cases: (1) optimization of liquid handling accuracy and (2) synthesis with computer vision monitoring. By bridging natural language interfaces, standardized protocols, and SDL interoperability, this architecture lowers the barrier to entry for both domain scientists and developers, paving the way for more adoptable, scalable, and intelligent laboratory automation.

Submission Track: Paper Track (Short Paper)

Submission Category: Automated Synthesis + Automated Material Characterization

Institution Location: Vancouver, Canada

Submission Number: 123