Due to the small size and to the high level of labour division, small firms in Industrial Districts need to establish horizontal and vertical cooperative relationships (Camuffo and Grandinetti, 2011; Malberg and Maskell, 1999) in order to share and combine complementary knowledge assets. Whilst the network approach (Granovetter 1985; Powell, 1991) has greatly contributed to the understanding of knowledge exchange processes in IDs and of their influence on performances, research on small firms’ networks has mostly assumed the network structure as a given. According to this gap, the purpose of the paper is to answer to the following research questions: is knowledge complementariness among firms in an industrial district a sufficient condition to let supply networks to emerge? What is the effect of relational embeddedness in determining the structural properties of these networks? We adopt an approach grounded on complexity science and consider the Industrial District as a Complex Adaptive System (Holland, 2002). The methodology used in the research is the agent-based simulation. We present an agent-based model of a stylized ID and build on it a virtual laboratory in which we perform generative experiments (Epstein and Axtell, 1996), in order to answer to the above research questions. In the literature on firms’ networks, topological works study how specific network structure influence the intensity of knowledge flows among a network’s firms. Our perspective is dual to the topological one: by not assuming that links among firms are pre-existent our objective is to generate the network topology with the help of an agent-based computational laboratory. Our contribution would be on different aspects: 1) we would explore how knowledge exchange processes can generate the emergence of network structures; 2) we analyse this topic in the context of small firms’ clusters, taking into account the crucial role that social characteristics of these systems play in shaping the phenomenon of network emergence; 3) the paper aims at contributing to the literature stream of dynamic emergence and evolution of supply networks, with a specific focus on the impact of collaborative strategies of agents on the emergent structures of networks. Main results of simulation experiments show that for every experimental set a stable network of links emerges among firms of the simulated ID. In addition, through the generative experiments we are able to identify certain conditions under which the emerged networks exhibit a hub&spoke structure. The model here proposed is not a case-based model, but an ideal-typical computational model, aimed at exploring and identifying a micro-macro relationship that could be applied to a class of empirical cases. As a consequence, the present research does not request, at this stage, a strict relation between the model and the empirical reality. The simulation is devoted to produce research hypotheses to be tested further using traditional methodologies. As this empirical validation process will be completed, the computational laboratory that has been presented in this paper could be used as a tool to support policy analysis e policy making decision processes.