Research on agents, intelligent agents, and multiagent systems (MAS) has been one of the most lively areas in the Artificial Intelligence (AI) field in the last years. Moving from basic theoretical and technological issues, MAS research has gone beyond the limits of classical (distributed) AI, spreading to apparently distant fields like software engineering, pervasive computing, self-organisation, simulation—essentially dealing with complexity of artificial systems, according to most of the legitimate acceptations of the term. In this paper, we devise out the main development lines of MAS research, and put them in historical perspective by focussing on the last decades of research in Italy.