You could think of the Anomaly Detector (also called anomalies) and clusters as similar models, since both models are unsupervised learning and both provide unlabeled data, however their learning task is different. Clusters group data by similarity whereas the anomalies assign a value from 0 (similar) to 1 (dissimilar) to each instance, therefore the closer to 1 the more anomalous that instance will be.
The most popular use cases for Anomaly Detection are fraud analysis, data cleaning, intrusion detection and authentication among others. Please watch here the webinar about Anomaly Detection and read its slides for a more detailed explanation.