Key points are not available for this paper at this time.
The convergence of generative AI with the emerging agentic AI paradigm, the Model Context Protocol (MCP), and urban digital twins offers transformative potential for automating the management of complex urban systems. In this paper, we demonstrate how agentic digital twins can generate knowledgeaugmented workflows through multi-step reasoning and autonomously execute them via coordinated AI agents equipped with scientific tools. These agents can discover and retrieve data, construct analytical pipelines, interface with domain-specific software, and adapt to dynamic urban conditions and userdefined scenarios through reflective reasoning. Drawing on agentic software for personal and recreation purposes, as well as prototype systems developed in the operations research domain, we illustrate how this paradigm enables both knowledge-augmented problem solving and automated orchestration of scientific tools to support real-time simulation and optimization. We further envision its broader potential for urban operations, including building energy management, traffic optimization, disaster response, and livability enhancement. Finally, we 1 outline a research agenda and key challenges for integrating agentic AI into urban digital twins, highlighting a pathway toward fully automated, user-interactive approaches for complex, interdisciplinary city management at scale.
Xu et al. (Wed,) studied this question.