The article introduces Gaia, a methodology for agent-oriented analysis and design. Gaia is designed to be both general and comprehensive, applicable to a wide range of multi-agent systems and addressing both macro-level societal aspects and micro-level agent interactions. The methodology is based on the concept of a multi-agent system as a computational organization of interacting roles. The authors argue that existing software development techniques, such as object-oriented analysis and design, are not well-suited for agent-based systems due to their fundamental mismatch with the agent-oriented view. Gaia aims to address this gap by providing a tailored approach to analyzing and designing agent-based systems. The article outlines the scope of Gaia, its conceptual framework, and its application in agent-based analysis and design. A case study on an agent-based business process management system is presented to illustrate the practical use of Gaia. The methodology is intended for large-scale real-world applications like ADEPT and ARCHON, where agents are coarse-grained computational systems with significant computational resources.The article introduces Gaia, a methodology for agent-oriented analysis and design. Gaia is designed to be both general and comprehensive, applicable to a wide range of multi-agent systems and addressing both macro-level societal aspects and micro-level agent interactions. The methodology is based on the concept of a multi-agent system as a computational organization of interacting roles. The authors argue that existing software development techniques, such as object-oriented analysis and design, are not well-suited for agent-based systems due to their fundamental mismatch with the agent-oriented view. Gaia aims to address this gap by providing a tailored approach to analyzing and designing agent-based systems. The article outlines the scope of Gaia, its conceptual framework, and its application in agent-based analysis and design. A case study on an agent-based business process management system is presented to illustrate the practical use of Gaia. The methodology is intended for large-scale real-world applications like ADEPT and ARCHON, where agents are coarse-grained computational systems with significant computational resources.