This paper proposes a novel neural-MCDM-based routing protocol, called NMRouting, to transmit data packets in mobile ad hoc network (MANET). Data packets are transmitted through neighbouring nodes that have high success rates from among a list of available neighbours. A perceptron-based neural controller specifies a success rate for every neighbour based on remaining energy, occupied buffer, and mobility speed. A MCDM controller determines the number of paths based on packet priority rate, distance, and average mobility speed. An energy management model is suggested to conduct energy consumption of the nodes efficiently as well as three queue types - critical, reliability, and ordinary - are considered to guarantee network reliability. Besides, the paper presents UML diagrams to develop NMRouting in real applications. Simulation results show that the proposed protocol has a high performance in terms of packet delivery ratio, network lifetime, and network traffic, compared to DSR, AODV, DPG and QMRPRNS.