Traffic control through semaphores, the so-called traffic lights, is an important asset at hand for engineers and practitioners in charge of managing large-scale urban networks. Optimization of traffic flow through city arteries depends highly on the way traffic lights are coordinated, and making them sort of intelligent in some way seems to be an excellent natural ground for the application of the autonomous agents metaphor. This paper reports on our preliminary work using reactive agents to test such an approach. Our main contribution relies on integrating an agent-based API layer to enhance an open-source simulator (SUMO, in this paper) to support a MAS-based traffic control platform to devise and test intelligent control in urban network scenarios. Some results are presented and discussed, as well as are some future work suggested.