In this study, a multi-agent receding horizon control is proposed for emergency control of long-term voltage instability in a multi-area power system. The proposed approach is based on a distributed control of intelligent agents in a multi-agent environment where each agent preserves its local information and communicates with its neighbours to find an optimal solution. In this study, optimality condition decomposition is used to decompose the overall problem into several subproblems, each to be solved by an individual agent. The main advantage of the proposed approach is that the agents can find an optimal solution without the interaction of any central controller and by communicating with only its immediate neighbours through neighbour-to-neighbour communication. The proposed control approach is tested using the Nordic-32 test system and simulation results show its effectiveness, particularly in terms of its ability to provide solution in distributed control environment and reduce the control complexity of the problem that may be experienced in a centralised environment. The proposed approach has been compared with the traditional Lagrangian decomposition method and is found to be better in terms of fast convergence and real-time application.