Cloud computing enables a revolutionary paradigm of consuming ICT services. However, due to the inadequately described service information, users often feel confused while trying to find the optimal services. Although some approaches are proposed to deal with cloud service retrieval and recommendation issues, they would only work for certain restricted scenarios in dealing with basic service specifications. Indeed, the missing extent is that most of the cloud services are "agile" whilst there are many vague service terms and descriptions. This paper proposes an agility-oriented and fuzziness-embedded cloud service ontology model, which adopts agility-centric design along with OWL2 (Web Ontology Language) fuzzy extensions. The captured cloud service specifications are maintained in an open and collaborative manner, as the fuzziness in the model accepts rating updates from users on the fly. The model enables comprehensive service specification by capturing cloud concept details and their interactions, even across multiple service categories and abstraction levels. Utilizing the model as a knowledge base, a service recommendation system prototype is developed. Case studies demonstrate that the approach can outperform existing practices by achieving effective service search, retrieval and recommendation outcomes.