The Dawn of the AI Co-Scientist: A Comprehensive Survey of Biomedical and Chemical Multimodal Agents and Their Applications

Autonomous and multimodal AI agents have emerged as transformative tools in biomedical and biochemical research. These AI coscientists are capable of autonomously performing a range of complex tasks in the scientific research lifecycle, while interacting with diverse modalities (e.g., molecular structures, biomedical images, omics sequences, and structured clinical records). This survey presents the first systematic review of over 80 papers on multimodal agentic AI in biomedical and chemical research. We propose a taxonomy of biomedical and chemical multimodal agents, covering model modalities, multimodal agent learning, multimodal agent inference, existing applications, and the current landscape of benchmarks and metrics. We provide a detailed review of representative approaches and datasets in each category. This survey serves as a valuable resource for researchers interested in interdisciplinary collaboration within biomedical and chemical agentic workflows.