Metaheuristic algorithms have been a very important topic in computer science since the start of evolutionary computing the Genetic Algorithms 1950s. By now these metaheuristic algorithms have become a very large family with successful applications in industry. A challenge which is always pondered on, is finding the suitable metaheuristic algorithm for a certain problem. The choice sometimes may have to be made after trying through many experiments or by the experiences of human experts. As each of the algorithms have their own strengths in solving different kinds of problems, in this paper we propose a framework of metaheuristic brick-up system. The flexibility of brick-up (like Lego) offers users to pick a collection of fundamental functions of metaheuristic algorithms that were known to perform well in the past. In order to verify this brickup concept, in this paper we propose to use the Monte Carlo method with upper confidence bounds applied to a decision tree in selecting appropriate functional pieces. This paper validates the basic concept and discusses the further works.