In this paper, we introduce a framework for modular generation of context-aware recommendations. The components of this framework include context sensors, recommender algorithms and utility modules (converters and filters), all realized as so-called services, which can flexibly be combined in terms of a recommender construction kit. Different areas of an application (e. g. a web portal) thus can be powered by a distinct recommendation-providing service chain. In this way, different recommender techniques can be used in parallel, either separately or combined, with or without incorporating contextual information, which makes it a framework for context-aware, hybrid recommendation generation.