The dynamics of MEMS 3D fluidic self-assembly (FSA) was modeled using interactive software agents, i.e. by agent-based modeling (ABM). ABM enables realistic simulations of 3D FSA dynamics taking into account spatial parameters - hard to include in analytic models. Our ABM model was tested by reproducing the experimental data of Zheng and Jacobs's 3D FSA process, and it was used to investigate the influence of design parameters and assembly strategies on FSA yield. The ABM model is a significant advance in the modeling of FSA and may represent the natural framework to explore open issues in this promising field.