Specialty Brazilian Canephora coffees are produced in the Amazon by indigenous and non-indigenous people and in Espírito Santo. Their distinctive quality, origin, and varietal were verified by integrated analytical techniques to understand better their chemical and sensory aspects and protect their origin and traceability. In this context, chemometric multi-block approaches represent a holistic way to integrate the multi-source data and then extract their complementary information. The samples were analyzed by near-infrared (NIR) spectroscopy on portable and benchtop instruments, ultraviolet-visible (UV-Vis) spectroscopy, mass spectrometry, and sensory analysis. Each piece of relevant information was interpreted before being integrated through exploratory data analysis and predictive modelling by multi-block methods. Although subtle, ComDim analysis showed a tendency to separate the indigenous and non-indigenous, and Espírito Santo coffees. Pre-processing ensembles with ROSA calibration (PROSAC) discriminated the three coffees with mean correct classification rate of 91.1 % in the test, using benchtop NIR with 1st derivative, mass spectrometry water spectra with Pareto scaling, and autoscaled sensory data. Sequential and orthogonalized partial least square-linear discriminant analysis (SO-PLS-LDA) performed better than PROSAC, showing 94.2 % of recognition in the test, using benchtop NIR with standard normal variate, mass spectrometry organic spectra with Pareto scaling, and portable NIR with 2nd derivative. Integrating complementary information from different blocks also improves classification accuracy compared to analyzing individual matrices.