The aim of this paper is to propose a general approach to include automated decomposition techniques in the mathematical programming modeling language. The motivation for such an approach has emerged from the following statement. On one side, modelers often choose to formulate their problems in well tested general purpose modeling languages such as GAMS and AMPL since the syntax of writing a model in this environment closely resembles the mathematical notation when written on paper. On the other side, cutting edge decomposition techniques are available to efficiently handle large scale mathematical programs possessing certain structure. The lack of a well developed link between these algorithms and the modeling languages prevents many modelers from accessing a rich source of algorithmic techniques.A modeling language is always linked with a solver through a set of functions that facilitate the communication between them. We propose to implement the automated decomposition at the level of these functions, with two additional functions. The first function can split the model data into separate data for the decomposed problems. The second function can take the separate problems and implement the decomposition algorithm by successively calling the solver to solve the separate decomposed problems until there is convergence to the optimal solution. This method has been currently implemented with two subroutines added in the GAMS I/O Library