Chronic cough is common, but automatic cough-monitoring technologies are not yet widely used in clinical practice. A major barrier is the lack of consistency in how cough events are defined, labelled and evaluated. Existing datasets are often private or heterogeneous, while different research groups use different performance metrics, making algorithms difficult to compare. ROAR (Respiratory Open Access Research) addressed these limitations by developing an open, standardized framework for cough-detection research. We established reproducible cough-event definitions and annotation principles, a standardized data structure, and an event-based methodology for evaluating automatic cough-detection algorithms. Previously acquired multimodal cough data were organized according to this framework, and open-source software was developed to support data handling and performance evaluation. The resulting resources - definitions, an open cough-counting dataset, validation methodology, benchmark metrics and software tools - were centralized on the ROAR website. This provides researchers with a common technical reference for developing, evaluating and comparing cough-detection systems reproducibly. The project also generated substantial community engagement. ROAR was discussed with international experts through a CHUV meeting and follow-up online meetings and was presented to the cough community at the European Respiratory Society (ERS) congress in 2025. These exchanges opened a pathway toward a future international consortium and potential ERS guidelines, an activity beyond the original project scope. ROAR will continue as an open research infrastructure maintained at EPFL and is intended to expand through additional datasets, algorithms and community contributions.