Demo: Orchestrating Large Language Model Agents and Resources for Medical Deep Research

Yuan Li, Matthew Pan, Claire Liu, Hengjin Zhu, Xijing Wang, Yingtao Luo

Keywords: medical, clinical, Deep Research, Multi-agent system

Abstract: Deep Research represents the convergence of large language models (LLMs), advanced reasoning, and information retrieval for expert-level inquiry. Existing Deep Research systems are constrained by reliability issues, limited integration with specialized resources, and inflexible output formats. In this paper, we introduce \textbf{Medical Deep Research}, an open, agentic system designed for in-depth medical and clinical investigations. Our framework features a multi-agent research module and a resource module that integrates curated medical tools and can dynamically discover Model Context Protocols (MCPs). Through fine-grained design for planning, tool orchestration, query processing, report formatting, and MCP integration, the system supports comprehensive medical information retrieval and customizable output generation. We evaluate this system across four key aspects: completeness, tractability, correctness, and helpfulness. Our evaluation results demonstrate the potential of Medical Deep Research to serve as a reliable and powerful platform for medical research. Code available at \url{https://github.com/Clinical-Copilot/Medical_Deep_Research}

Submission Number: 149