Agentic Information Architectures for Global Climate Governance

Global climate governance depends on information systems that turn fragmented national policies into interoperable knowledge. Existing repositories are largely static and descriptive, which limits evidence-based learning across jurisdictions. The authors present CAPAS, a Cross National Agent Based Policy Analysis System, that integrates large language models, a policy ontology, and multi agent orchestration for automated extraction and recommendation. CAPAS structures 70,000 mitigation policies into seven dimensions covering instruments, actors, sectors, targets, timelines, and implementation mechanisms, enabling fine grained comparison and semantic alignment. A LangGraph-based workflow classifies, extracts, and recommends analogous policies with transparent reasoning. Experiments show improvements in interpretability and transferability. From an information governance view, the study shows how agentic architectures operationalize the information lifecycle across jurisdictions and extends decision support system theory to global policy analysis.