Theoretical trends in agent-based modeling (ABM) draw sharp lines that usually limjt the expressiveness of models to fit their methodological box: KISS puts emphasis on parsimonious and tractable model for systematic simulation analysis, KIDS focuses on the use of data for model specification and simulation validation, while KILT highlights the involvement of stakeholders to build representative model and turn simu lation into a learning tool. In this proposal, we stress the benefit to break the lines and reinstate the various agent-based model purposes, focuses and supports in the particular context of designing transformative ABM. The proposed perspective to burst methodological boxes is based on the creation of models to mix social actors and issues based on sub-models supporting a variety of theoretical approaches. To back this methodological claim, we detailed a I O years-long modeling effort to represent and support community based management of renewable resources in the wetlands of the lower Amazon, in the Para state of Brazil.