Constrained objects provide a suitable object-oriented style for modeling systems under constraints. A set of classes is defined to represent a problem, whose state is then controlled by a constraint satisfaction engine. This engine is commonly a black-box based on a predefined and non-customizable search strategy. This system rigidity, of course, does not allow users to tune models in order to improve the search process. In this paper we target this issue by presenting an extensible formalism to define a wide range of search options so as to customize, improve and/or analyze the search process of constrained object models. © 2008 Springer-Verlag Berlin Heidelberg.