This study presents an algorithm that determines the optimal configuration of the FMSs with limited local buffers. The algorithm finds the lowest cost configuration, i.e. the number of tools, the number of pallets as well as the number of buffers to be installed in front of each machine in the system. Thus it assures a given production ratio with a minimum cost. In the algorithm, FMSs are considered as the closed queueing network with limited queue length. System performance evaluation is performed using the Block-&-Recirculation model developed by Yao and Buzacott. The algorithm is composed with three steps. The steps are namely i) determination of a lower configuration, ii) derivation of an heuristic solution, and iii) obtaining the optimal solution. The computational efforts required in the algorithm usually lies within the capability of personal computers.