For the last decade, teams at Oracle relied on the Spoofax language workbench to develop a family of domain-specific languages for graph analytics in research projects and in product development. In this paper, we analyze the requirements for integrating language processors into large-scale graph analytics toolkits and for the development of these language processors as part of a larger product development process. We discuss how Spoofax helps to meet these requirements and point out the need for future improvements.