Machine learning algorithms can be used to combat fake news propagation. For the news classification, labeled datasets are required, however, among the existing datasets, few separate verified false from skewed ones with a good variety of sources. This work presents FACTCK.BR, a new dataset to study Fake News in Portuguese, presenting a supposedly false News along with their respective fact check and classification. The data is collected from the ClaimReview, a structured data schema used by fact check agencies to share their results in search engines, enabling data collect in real time.