Novel biomarkers are essential to improve the treatment efficacy and overall survival of stage II and III colorectal cancer (CRC), allowing for personalized treatment decisions. Here, the densities of CD8 and FOXP3 T cells in the tumor and invasive margin were processed by immunohistochemistry and digital pathology to form a scoring system named regulatory-Immunoscore (RIS). Cox proportional hazards regression models were used to determine the risk factors associated with time to recurrence. Harrell's concordance index and the time-dependent area under the curve were used to assess model performance. A total of 1213 stage I-III DNA mismatch repair-proficient colorectal cancer (pMMR CRC) patients were randomly assigned to a training set (n = 642) and a validation set (n = 571). From the Cox multivariable analysis, the association of RIS with survival was independent of patient age, sex and anatomy-based tumor risk parameters (P < .0001). For stage II patients, chemotherapy was significantly associated with better recurrence time in patients with low (95% confidence interval [CI]: 0.11-0.54, P = .001) and intermediate (95% CI = 0.25-0.57, P < .001) RIS values. In stage III patients treated with adjuvant chemotherapy, a treatment duration of 6 or more months was significantly associated with better recurrence time in patients with intermediate RIS values (95% CI = 0.38-0.90, P = .016) when compared with duration under 6 months. Therefore, these findings suggest that RIS is reliable for predicting recurrence risk and treatment responsiveness for patients with stage I-III pMMR CRC.