The traditional way of constructing items to assess learning is time-consuming because it requires experts to perform the labor-intensive task of creating legions of items by hand. The Automatic Item Generation (AIG) approach aims to streamline this process by having experts not formulate individual items, but rather create highly structured models that can be used by software to automatically generate them. This requires two types of software components. First, an editor that allows experts to specify these models. Second, a generator that processes the specified models and generates the intended items. The elaboration of these components must address the following challenges. The former must be usable for the definition of complex knowledge models while the corresponding modeling process should be easy to understand. The latter should be able to process these models in a reasonable time. Thus, the goal is to overcome both challenges by defining and conceptualizing the use of a model representation that is easy to understand and efficient to process. Therefore, we present a new AIG software toolkit in relation to our previous work, which addresses these challenges by introducing a new representation approach - the layered-model-approach. The toolkit shall be evaluated in terms of usability and efficiency.