As domain-specific modeling languages, and especially graphical modeling languages become more and more widespread, so do workbenches for generating editors and tools to use said languages. The advantages of these languages depend heavily on their expressiveness and suitability to communicate stakeholders' concerns; therefore, new languages have to be continuously developed, frequently on top of existing languages, by reusing concepts and icons. Unfortunately, many of these new languages have to be constructed from scratch because it is difficult to adapt, compose, and reuse the tools previously created for the base languages. In this paper we present Gromp, a language and a platform that improves this situation. Gromp was built on top of Picture, our platform for the description of graphical syntaxes and generation of editors. The core of Gromp is a language that offers the mechanisms to describe the composition of multiple graphical languages, and to generate the corresponding editors. This results in languages that can be designed in a more modular way and be easily adapted and reused.