Introduction. There has been a growing interest towards cognitive-training programmes to improve cognition and prevent cognitive impairment despite discrepant findings. Physical activity has been recognised in maintaining or improving cognitive ability. Based on a psychoneurophysiological approach, physiological indexes should partly determine neuronal dynamics and influence cognition as any effects of cognitive training. This study’s primary aim was to examine if improved physiological indexes predict improved cognitive variables in the context of a clinical intervention programme for type 2 diabetes (T2D). Method and analysis. PhyCog will be a 22-week randomised controlled trial comparing cognitive performance between three arms: (1) physical activity (1 month), a 15-day wash-out, then cognitive training (1 month), (2) cognitive training (1 month), a 15-day wash-out and physical activity (1 month), and (3) an active breathing condition (psychoeducation and resonance frequency breathing for 1 month), then a 15-day wash-out, and combined physical activity and cognitive training (1 month), allowing to determine the most effective intervention to prevent cognitive impairment associated with T2D. All participants will be observed for 3 months following the intervention. The study will include a total of 81 patients with T2D. Cognitive performance and physiological variables will be assessed at baseline (week 0—W0), during the washout (W5, 72–96 hours after week 4), at the end of the intervention (W10), and at the end of the follow-up (W22). The main variables of interest will be executive function, memory and attention. Physiological testing will involve allostatic load such as heart rate variability, microcirculation, cortisol and dehydroepiandrosterone sulfate levels. Sociodemographic and body composition will also be a consideration. Assessors will all be blinded to outcomes. To test the primary hypothesis, the relationship between improvement in physiological variables and improvement in cognitive variables (executive, memory and attention) will be collected. Ethics and dissemination. This protocol was approved by the Est III French Ethics Committee (2020-A03228-31). Results will be published in peer-reviewed journals. Trial registration number NCT04915339.