A fair number of epidemiological studies suggest that age at menarche (AAM) is associated with depression, but the reported effect sizes are small, and there is evidence of residual confounding. Moreover, previous Mendelian randomization (MR) studies to avoid inferential problems inherent to epidemiological studies have provided mixed findings. To clarify the causal relationship between age at menarche and broadly defined depression risk, we used 360 genome-wide significantly AAM-related single-nucleotide polymorphisms (SNPs) as instrumental variable and data from the latest GWAS for the broadly defined depression risk on 807,553 individuals (246,363 cases and 561,190 controls). Multiple methods to account for heterogeneity of the instrumental variable (penalized weighted median, MR Lasso, and contamination mixture method), systematic and idiosyncratic pleiotropy (MR RAPS), and horizontal pleiotropy (MR PRESSO and multivariable MR using three methods) were used. Body mass index, education attainment, and total white blood count were considered pleiotropic phenotypes in the multivariable MR analysis. In the univariable [inverse-variance weighted (IVW): OR = 0.96, 95% confidence interval = 0.94–0.98, p = 0.0003] and multivariable MR analysis (IVW: OR = 0.96, 95% confidence interval = 0.94–0.99, p = 0.007), there was a significant causal effect of AAM on depression risk. Thus, the present study supports conclusions from previous epidemiological studies implicating AAM in depression without the pitfalls of residual confounding and reverse causation. Considering the adverse consequences of an earlier AAM on mental health, this finding should foster efforts to address risk factors that promote an earlier AAM.