Recommender systems (RSs) are complex software systems that suggest relevant items of interest given a specific application domain to users. The development of RSs encompasses the execution of different steps, including data preprocessing, choice of appropriate algorithms, item delivery, to name a few. Though RSs can alleviate the curse of information overload, existing approaches resemble black-box systems, where the end-user is not supposed to customize the overall process. To fill the gap, we proposed LEV4REC, an initial MDE-based prototype for supporting the mentioned activities needed to conceive an RS from the design phase to the actual deployment of the system, including the parameters fine-tuning. As a first step, the user defines a coarse-grain model that allows the configuration of the desired RS, which then can be finalized by fine-tuning different parameters. LEV4REC eventually generates the source code of the RS, being ready for actual deployment. LEV4REC is provided as a plugin extension for two different IDEs and an initial web interface. To study the capabilities of the approach, we utilized LEV4REC in curating two existing RSs, which have been designed on top of two different algorithms, i.e., collaborative filtering and feed-forward neural network. Experimental results show that our proposed tool can create systems that can provide suitable recommendations, thereby conforming to their original design. LEV4REC is capable of enabling developers to refine the produced system by experimenting with different algorithms, experimental settings, and evaluation metrics.