In the past twenty years we have seen an enormous growth and development of new techniques, technologies, and tools that support the engineering of Multi-Agent Systems (MAS). The 1990s perhaps are best characterized as the period in which the foundations were laid and the more theoretical underpinnings of the MAS field were explored. Besides a continuation of this foundational work, since 2000 the agent-based community has also been increasingly able to demonstrate the great potential for applying the MAS technology that has been developed in a very broad and diverse range of application domains. In this paper, I will trace the shaping of the agent-oriented mindset from the mid 90s on as it evolved in the work presented in the international workshops ProMAS, AOSE, and DALT that recently merged into the EMAS workshop. For this reason the focus of this overview will be in particular on cognitive agents as it seems fair to say that most work reported in ProMAS, AOSE, and DALT has taken its inspiration from Belief-Desire-Intention (BDI) agents.