Abstract Background cumulative evidence from cohort studies suggested that there were inconsistent conclusions as to whether there was a bidirectional association between depression and frailty. Therefore, this study used a bidirectional two-sample Mendelian randomisation (MR) study to investigate the causal relationship between depression and frailty. Methods we performed univariate and multivariate bidirectional MR analyses to assess the causal association between depression and frailty. Independent genetic variants associated with depression and frailty were selected as instrumental variables. Inverse variance weighted (IVW), MR-Egger, weighted median and weighted mode were mainly used in univariate MR analysis. Multivariate MR (MVMR) analyses used multivariable inverse variance-weighted methods to individually and jointly adjust for three potential confounders, body mass index (BMI), age at menarche (AAM) and waist-to-hip ratio (WHR, adjusted for BMI). Results univariate MR analysis showed a positive causal relationship between depression and risk of frailty (IVW, odds ratio (OR) = 1.30, 95% confidence interval (CI) = 1.23–1.37, P = 6.54E−22). Causal relationship between frailty and risk of depression (IVW, OR = 1.69, 95% CI = 1.33–2.16, P = 2.09E−05). MVMR analysis revealed that the bidirectional causal association between depression and frailty remained after adjusting for three potential confounders, BMI, AAM and WHR (adjusted for BMI), individually and in combination. Conclusions our findings supported a causal relationship between genetically predicted depression and frailty in both directions.