In this paper we discuss a multi-agent cooperative approach for decentralized, coordinated traffic systems. Intersection controllers are equipped with agents, i.e., autonomous software systems which are capable of communicating and cooperating with one another to achieve an individual or global goal. Our approach is based on real-world traffic parameters and constraints, and is meant to be implemented in existing traffic systems with minimal changes. Experimental results show that the agent-based approach outperforms the traditional pre-timed and actuated modes when traffic is heavy.