Portable near-infrared spectroscopy devices allow measurements of muscle oxygen saturation (SmO2) in real time and non-invasively. To use NIRS for typical applications including intensity control and load monitoring, the day-to-day variability needs to be known to interpret changes confidently. This study investigates the absolute and relative test-retest reliability of the Moxy Monitor and investigates side differences of SmO2 at the vastus lateralis muscle of both legs in cyclists. Twelve trained cyclists and triathletes completed 3 incremental step tests with 5min step duration starting at 1.0W/kg with an increase of 0.5W/kg separated by 2-7days. SmO2 was averaged over the last minute of each stage. For all power outputs, the intra-class coefficient (ICC), the standard error of measurement (SEM) and the minimal detectable change (MDC) were calculated. Dominant and non-dominant leg SmO2 were compared using a three-factor ANOVA and limits of agreement (LoA). ANOVA showed no significant systematic differences between trials and side. For both legs and all intensities, the ICC ranged from 0.79 to 0.92, the SEM from 5 to 9% SmO2 and the MDC from 14 to 18% SmO2. The bias and LoA between both legs were -2.0% ± 19.9% SmO2. Relative reliability of SmO2 was numerically good to excellent according to current standards. However, it depends on the specific analytical goal whether the test-retest reliability is deemed sufficient. Wide LoA indicate side differences in muscle oxygenation during exercise unexplained by leg dominance.