Complex adaptive systems are decentralized, distributed, and dynamic in nature. Consequently, these systems cannot be analyzed successfully by modeling them as single cognitive agents since the capacity of an agent is limited by its knowledge and computing resources. Multi-agent modeling by its modular characteristic provides a valuable engineering abstraction for analysis of such systems. This paper focuses on multi-agent system architectures and design methodologies, as well as models of evolution integrated in these studies. As a case study of an agent-based architecture, an evolutionary financial market architecture is outlined.