Recent studies have shown that various models could explain the emergence of complex networks, such as scale-free and small-world networks. This paper proposes another novel model to generate complex networks using a multi-agent approach. Each node is considered as an agent, and based on voting by all agents, an edge is added newly. We use four different kinds of centrality measures as utilities of each agent. Depending on a utility to use, the resultant network differs significantly; typically closeness centrality generates a scale-free network, degree centrality generates a random graph, betweenness centrality rusults in a broad graph, and eigenvector centrality brings a complete subgraph. The paper illustrates the connection between local behavior of an agent and the global property as an entire network, which contributes various aspect of multi-agent applications.