Wireless cellular networks are used for data services like browsing, streaming, and conferencing. The increased demand for guaranteed service performance as well as for higher resources as required for bandwidth thirsty applications demand for an efficient call admission control mechanism. Most of the existing solutions targeting to meet this dipole demand follow either a predictive mechanism or some statistical approach. Predictive solutions require extensive computational load over switching centers. Further, the efficiency of solution depends on the degree of accuracy of the prediction. On the contrary, statistical approaches reduce the overhead to some extent through distributed mechanism. However, it needs synchronization among participating components. This paper presents an agent based distributed scheme for efficient management of radio resources in cellular networks such that QoS is assured for every transaction. Agents coupled to network components, like Radio Resource Manager, actively participate in network resource management. In the perspective of call admission control, agents facilitate proactive and reactive capturing requirement of each application's bandwidth. Agents also anticipate the resource requirement continually and optimize allocation of resources with fairness among the contending user applications. Besides, the proposed solution does not require any additional infrastructural and would perform better in terms of call blocking probability. © 2011 Springer-Verlag.