This paper describes a novel architecture for anti-terrorism planning and resource allocation systems. The architecture uses web services and agents to provide a re-usable and extensible architecture for these types of systems. A dynamic web-based client and custom middle ware provide for the integration of otherwise de-coupled software components. The architecture has evolved from a longitudinal field study of U.S. Marine Corps anti-terrorism planners through recent empirical work in a more controlled setting. We describe this evolution, the architecture and its rationale, and provide some lessons learned for others engaged in development of anti-terrorism systems.