In this paper we have adopted an agent approach for traffic light control. According to this approach, our system consists of agents and their world. In the traffic context, the world consists of cars, road networks, traffic lights, etc. Each of these agents controls all traffic lights at a road junction by an observe-think-act cycle. That is, the agent repeatedly observes the current traffic condition at the junction, it then uses this information to reason with condition-action rules to determine how the agent should act in what traffic condition, and finally it performs those actions in order to efficiently manage the traffic flows. We have also developed a NetLogo-based traffic simulator to serve as the agents? world. Our approach is experimented with traffic control of a few connected junctions and the result obtained is promising; it can reduce the average delayed time of each car at each traffic-light near a junction rather substantially when compared with other approaches.