The multi-agent paradigm provides models of autonomous, interacting entities capturing the complexity many of today’s software systems. The design of multi-agent systems remains a challenge, involving high project risk due to insufficient, late understanding of system behaviours and poor consideration of stakeholder requirements. Concepts from agile software development approaches, including incremental development and early prototyping, have been proposed to address these issues. Yet, most agent-oriented development approaches have not integrated these principles. In this paper, an incremental methodology for developing multi-agent systems is proposed based on the integration of subject-oriented process management (S-BPM) and a generalised approach for developing multi-agent systems. A toolchain has been implemented for demonstrating the methodology.