The importance of learner corpora for foreign language acquisition research as well as their role in data-driven learning and other learning contexts is now widely recognised. They have become a valuable resource for both foreign language teaching and learning. To date, there is no extensive collection of learner language data from L1 German speakers for the Romance languages taught in schools (French, Spanish, Italian). This desideratum is addressed by the error-annotated learner corpus MuLeCo (Munich Learner Corpus). The collection of written learner productions aims to shed light on persistent challenges faced by learners of French, Spanish, and Italian, while also providing a solid empirical basis for developing didactic and data-driven materials for foreign language teaching—thus helping to bridge the gap between Foreign Language Acquisition (FLA) and Foreign Language Teaching (FLT). In addition, MuLeCo creates a space for critically revisiting key concepts such as “error,” “variation,” and “norm” in the context of interlanguage phenomena. This article aims to demonstrate how a learner corpus can be constructed to identify persistent problem areas in foreign language learning and processing. Following an outline of the linguistic and didactic objectives, the article presents in detail the methodology employed to collect, structure, organise, analyse, and make the corpus data accessible. The entire workflow is designed to be fully digital. At the core of the corpus lies the categorisation of errors. The relational database used for storing and handling the highly structured corpus data allows for multifold analysis. The article concludes with initial analytical approaches and selected findings