Abstract: Artificial intelligence is reshaping the knowledge production, methodological organization, and governance mechanisms of urban planning. As urban systems become more complex, public-interest coordination more demanding, and planning implementation more uncertain, urban planning urgently needs a professional intelligent methodology that covers the whole planning process. This paper proposes the concept of the urban planning agent, defined as a professional intelligent system that uses multi-source urban data as its perceptual basis, draws on planning knowledge, regulations and standards, model-based methods, case experience, and implementation feedback, and makes judgments according to public value, spatial order, and governance objectives. The paper discusses the conceptual connotation, capability structure, system architecture, application map, and Chinese institutional context of urban planning agents. It proposes an eight-dimensional capability framework of perception, cognition, simulation, creation, decision-making, execution, coordination, and learning; constructs a four-layer operating architecture of data, computation, decision-making, and application; and identifies eight types of specialized agents: spatiotemporal sensing, regularity cognition, planning simulation, plan-making, action decision-making, organizational coordination, emergency response, and knowledge iteration. In China, urban planning agents should be embedded in the territorial spatial planning system and aligned with unified natural-resource management, urban assessment, urban regeneration, and digital government construction. Their operation should be supported by five databases covering natural resources, urban construction, socio-economic conditions, planning standards and regulations, and planning cases. Urban planning agents can help shift urban planning from stage-based plan preparation toward continuous spatial governance, offering theoretical and practical pathways for territorial spatial governance and human-AI collaborative planning.