This paper presents an algorithm inventory policies for a class of multi-product dependent setup cost problems. The operational policy under consideration is to replenish jointly, at a reduced cost, all products of the dependent class which are to be ordered in a scheduling period. Demands are assumed to be random with known probability distributions. Convergence of the iterative algorithm, to the optimal solution under suitable hypotheses is proven. Comparisons with the results of a simpler algorithm by Simmons demonstrate the cost advantages of our method.