Large language models (LLMs) and agentic AI are emerging as transformative technologies across geosciences, natural hazards, and disaster management. Their potential ranges from advancing risk knowledge and hazard monitoring to enhancing early warning systems, situational awareness, and operational decision-making. Yet, important challenges remain: reliability of outputs, data interoperability, transparency, and trust among operational stakeholders. This session invites contributions that explore conceptual, methodological, and applied aspects of LLMs and agentic AI in geosciences. We welcome case studies demonstrating their use in fields such as hazard forecasting, risk communication, and emergency response; research on retrieval-augmented generation, agent frameworks, and multi-modal approaches; and critical perspectives on ethics, governance, and the operational adoption of these tools. By bringing together scientists, technologists, and practitioners, the session aims to advance the discussion on how LLMs and agentic AI can bridge the gap between complex risk/climate knowledge and actionable insights for disaster management.