This paper considers an inventory-routing problem in a distribution network in which kanban is used as a means to implement just-in-time strategy. There are a set of part suppliers and cross-docks in this network, which provide parts for an assembly plant. A novel mixed integer non-linear programming formulation is developed for the problem in which the optimum number of kanbans is determined considering inventory and transportation cost. In the other word, this research aims at reducing the level of inventory and transportation cost in a kanban controlled supply chain system, in which the number of kanbans is to be determined optimally considering demand from suppliers and geographical distribution of them. A fleet of vehicles is applied to transport kanbans from suppliers to the assembly plant via two transportation strategies: direct shipment and shipment through cross-docks (indirect shipment). In the second strategy, it is possible to have routes between suppliers. The proposed problem is NP-hard based on literature, thus a memetic algorithm is introduced to solve it. Solving several examples reveals that the solving method significantly outperforms GAMS/CPLEX in reducing objective value and computational time.