In the international intelligence community, global repositories evolve naturally as a result of the multi-organizational collection of valuable information and insights. As individual analyst workspaces grow and links between analysts multiply, this global repository becomes a complex inter-related network of information. Analysts solve problems through collaboration at two levels: by sharing large quantities of new information and interacting directly with their colleagues to share insights. The authors are designing an infrastructure to enable analytical collaboration. An innovation in this work is a two-tiered model-based approach with automated, context-based information sharing on one level and person-in-the-loop insight sharing on the other. In evaluating our approach, the initial implementation uses a non-relational data store mapped into multiple viewpoints with a modeling language expressed in XML. This architecture applies to any system promoting cooperative analysis in a globally-distributed repository, not just systems in the intelligence community.