Introduction: The association between social capital and depression is a frequent research topic in developed countries, often with inconclusive results. Furthermore, for both social capital and depression, there are gender differences established in the literature. This study investigates gender differences in the association of social capital with the incidence and maintenance of depressive episodes. Methods: Baseline and second wave data (4 years of follow-up) from the Longitudinal Study of Adult Health (ELSA-Brasil), a multicenter cohort of civil servants with 15,105 workers aged 35–74 years, were used. Social capital was assessed using the Resource Generator, a scale composed of two different dimensions: “social support” and “prestige and education.” Depressive episodes were assessed using the Clinical Interview Schedule - Revised (CIS-R). The statistical analysis was performed using multinomial regression with adjustments for possible confounding factors. Results: Among men, low social capital in the “social support” dimension was associated with the incidence of depressive episodes (RR = 1.66; 95% CI: 1.01–2.72). Among women, social support was associated with the maintenance of depressive episodes (RR = 2.66; 95% CI: 1.61–4.41). Social capital was not associated with the incidence or maintenance of depressive episodes in the “prestige and education” dimension in both genders. Conclusion: The results highlight the importance of the dimension “social support” in both genders in its association with mental health. The resource-based social capital approach proved to be adequate for investigating mental health and confirms the idea that social networks can be useful in the treatment and prevention of depressive episodes.