Determining the on-tree maturity stage of mango fruits is essential for optimising harvest timing and maximising postharvest quality. This study investigates the feasibility of Near Infrared Spectroscopy (NIRS) as a reliable technology for establishing the optimal harvest time of mangoes. To achieve this, two portable near infrared (NIR) spectrophotometers —the MicroNIR™ Pro 1700 and the pre-calibrated F-751 Mango Quality Meter— with distinct optical designs were compared for in situ, on-tree analysis. A total of 210 mangoes from the ‘Osteen’ and ‘Keitt’ cultivars, were analysed. Predictive NIRS models were developed using the MicroNIR™ Pro 1700 to estimate skin colour, flesh colour, pulp firmness, dry matter (DM), and soluble solids content (SSC), employing modified partial least squares (MPLS) regression and various spectral pre-treatments. The F-751, with commercial calibrations based on other mango cultivars, was used solely to predict DM and SSC. The residual predictive deviation for cross-validation (RPDcv) values for the combined ‘Osteen’ and ‘Keitt’ sample set confirmed the efficacy of the MicroNIR™ in assessing flesh colour (L*) (RPDcv = 1.83), pulp firmness (RPDcv = 1.80), DM (RPDcv = 1.55), and SSC (RPDcv = 3.42), the main parameters used to establish the harvest index. These findings confirm the viability of using NIRS technology as a reliable decision-support tool for monitoring the on-tree evolution of mango internal quality and determining the optimal harvest time. However, the routine implementation of the F-751 pre-calibrated models proved unsuitable for accurately predicting key quality parameters under the mango growing conditions in Spain.