Today Online Learning Environments (OLE) like MOOCs and LMS are very commonly used and a huge number of students with very different profiles and backgrounds follow the same online courses. Still, personalized experience for attendees is not widely spread on the platforms hosting these courses. At the same time, there is a growing number of open educational resources (OER) that can helpfully enrich the content of online courses and even be chosen to match one-by-one the student tastes. Recommender systems may support personalization in OLE by providing each learner with learning objects carefully selected to help reaching their learning objectives. This kind of recommendation is more specific to compute than usual recommendations like consumer products: the recommendation depends not only on the learner profile, but also on the content of the course, because the recommendation needs to fit precisely with the course format at any point. In this article, we introduce MOORS, a MOOC-based OER recommender system that can be plugged in an OLE to dynamically provide recommendations of OER to learners on the basis of their profiles and the profile of the MOOC. We also describe the process for calculating recommendations from OER metadata, assuming these metadata follow the Linked Opend Data (LOD) principles. Our implementation has been done in Open edX, an open source MOOC platform widely used, however the same approach could be implemented in any OLE as long as the learners profiles and the course profile can be extracted. Finally, we discuss a life-size evaluation of our recommender system.