Background: Poverty is a social determinant of poor mental health outcomes. Agricultural interventions may reduce the burden of poor mental health outcomes for depression, and stress by enabling economic pathways out of poverty, particularly when paired with a group-based microfinance programs that provide complementary economic support. We estimated the relationship between an agricultural intervention, depression and psychological stress outcomes. Methods: We conducted a cross-sectional survey using validated depression and stress scales among 312 participants. The exposed group received farm inputs and agribusiness trainings. We fit an adjusted multivariate linear regression models to estimate the association for depression and stress scores, in addition to a quantile regression model to examine variation of effect sizes across depression score quantiles. We tested for effect measure modification by wealth status as measured with a composite score derived from 20-household and agricultural assets. Results: The agricultural intervention was associated with lower depression ({beta} = -1.81; 95% CI: -4.24, 0.62) and stress scores ({beta} = -1.39; 95% CI: -3.35, 0.56), but estimates were imprecisely measured. The association was strong in those with highest depression scores (Q90: {beta} = -7.30; 95% CI: -13.59, -0.99), and weaker in those with lower depression scores (Q10: {beta} = -0.7; 95% CI: -2.93, 1.48). There was no effect measure modification by wealth status (Wald p = 0.24), although larger effect sizes for depression were observed in those with low household wealth ({beta} = -5.23; 95% CI: -10.75, 0.33). Conclusion: An agricultural intervention was associated with modestly lower depression and stress scores, although overall estimates were imprecise. The association was strongest among individuals with severe depression, suggesting agricultural support may be beneficial for those experiencing greater psychosocial burden. Integrating agricultural support into microfinance programs may reduce depression among individuals with greater need. Future longitudinal work with larger samples may improve our causal understanding of these relationships.