We present an agent-based approach to Dantzig-Wolfe column generation (a decomposition approach), which is applied to an integrated production, inventory, and distribution routing optimization problem. The decomposition model has been implemented in the Java programming language, using the Java Agent DEvelopment Framework (JADE) and the ILOG CPLEX mixed integer linear optimization problem solver. The model has been validated oil a set of realistic scenarios and based on the results, we estimate the potential performance improvement which can be obtained by using a completely distributed implementation. We analyze the overhead, in terms of communication costs, that is imposed by an agent-based approach.