|Current studies on the intelligent distributed framework in distributed systems use multi-agents which include replication agents, grouping agents, locator agents, and load balancing agents among others which work systematically to provide quality of service. This research aims to improve that quality of service by implementing a neural network in the fault tolerant scheme. Incase of an object failure or disconnection, the properties of that object will be used as input data in the multilayer perceptron (MLP) to select an .alternate object. The fault tolerant scheme then chooses a new object by training these properties using the back propagation algorithm. Results show that the proposed algorithm recorded the highest accuracy rate compared to the ZeroR, Simple Logistic, and J48 algorithms.