MCP-Enabled LLM Agents for Closed-Loop Optimization in Real-Time Physical Experiments

zekun ren, Yee Jia En, Tan Hong Zhao, Qianxiao Li, Kedar Hippalgaonkar

Submission Type: I want my submission to be considered for both oral and poster presentation.

Keywords: Model Context Protocol (MCP), self-driving laboratory, closed-loop optimization, large language models, multi-agent systems, laboratory automation, Bayesian optimization, electrochemistry

TL;DR: We built an MCP-based multi-agent lab platform where LLMs run real experiments in closed loop; across physical benchmarks, LLM-driven optimization outperforms Bayesian optimization, and we extend the workflow to electrochemistry optimization.

Confirmation Of Submission Requirements: I submit an abstract. It uses the template provided on the submission page and is no longer than 2 pages.

Submission Number: 107