A number of IoT devices are continuously increasing day by day and these devices are delay-sensitive and require a quick response. These devices are connect-ed to a cloud environment for processing but there is latency in computation. To overcome this situation, the fog environment is kept at the edge of the IoT device layer to perform the quick computation for delay-sensitive applications. For quick computation, adequate number of resource should be available for processing of request. Resource provisioning is big challenge in fog computing and hence intelligent technique for resource provisioning is required. Here in this paper, we pro-posed an multi agent based algorithm for resource provisioning which allocate re-sources according to load. This paper consider three types of agents for resource provisioning e.g. task agent, fog agent and node agent. To evaluate the performance of proposed agent based resource provisioning we have compared it with already proposed technique of resource provisioning and simulation results demonstrates that proposed strategy minimizes makespan, average execution time and flowtime as compared to existing resource provisioning technique by Meera et al. [1].