The challenge we address in this work is the control of complex systems. We consider techno-social systems which share with complex systems some characteristics like big number of entities, autonomous entities, pre-existing systems, and multiple levels of organization. Our proposal is based on the multi-agent paradigm to model the complex system, to forecast its evolution and to assess the impact of control actions on it. Our solution is an exogenous architecture of the system to control using an equation-free approach based on multi-agent model. An example implementation of the architecture is presented on a free-riding problem of peer-to-peer le sharing networks. Implementation results demonstrate that our architecture can control such a network. Our contributions are i) to demonstrate the feasibility of our approach and its ability to control a system and, ii) to show that the modularity of the architecture enables to tackle issues related to the use of multi-agent paradigm in the context of control of complex systems.