Background: Although sedentary behavior is associated with the onset of major depressive disorder, it remains unclear whether sedentary behavior at work increases the risk of depression. The present study used the Bayesian approach to investigate the association between sitting time at work and the onset of major depressive episode (MDE). Methods: A 1-year prospective cohort study was conducted among 233 Japanese workers without MDE (response rate: 4.3%). MDE onset was assessed using the self-reported WHO Composite International Diagnostic Interview version 3.0. A Bayesian Cox proportional hazard model was used to estimate the hazard ratio (HR) between long sitting time at work and MDE onset. Results: A total of 231 workers were included in the analysis. During the follow-up, 1621 person-months were observed, and six participants experienced MDE onset. Incident rates per months were 0.34, 0.11, and 1.02% in short (< 7.2 h per day), medium (7.2–9.5 h), and long (9.5+ h) sitting time at work, respectively. The estimated median posterior probability distribution of the HR of long sitting time was 3.00 (95% highest density interval [HDI]: 0.73–12.03). The estimated median remained positive after adjustment for physical activity level and other covariates (HR = 2.11, 95% HDI: 0.42–10.22). The 10-base Bayesian factor for H1 (HR = 1.00) compared with the alternatives (H0, HR = 1.00) was 0.68 in the adjusted model. The analysis, which treated sitting time at work as a continuous variable, estimated that the median of the posterior probability distribution of the HR of sitting time was 0.79 (95% HDI: 0.58–1.07. The 10-base Bayesian factor was 2.73 in the linear association. Conclusions: Long sitting time at work (9.5+ h per day) might be associated with MDE onset among workers. However, the linear association indicated conflicting results. Non-linear associations between sitting time and MDE onset might explain this inconsistency. The evidence for an adverse association between sitting time at work and MDE onset remains inconclusive.