Restless legs syndrome (RLS) has a multifactorial etiology, and current treatments are suboptimal. Micronutrients influence neuromuscular and dopaminergic function, yet their causal role in RLS is uncertain. This study aimed to investigate whether circulating micronutrients causally influence the risk of RLS by applying an integrated two-sample, bidirectional, and multivariable Mendelian Randomization (MR) strategy. Genetic instruments for micronutrient levels were obtained from genome-wide association studies (GWAS) in the IEU Open GWAS database. RLS outcome data were from the EU-RLS-GENE consortium (7248 cases, 19,802 controls of European ancestry). Single nucleotide polymorphisms significantly associated with each micronutrient ( < 5 × 10) served as instrumental variables. Primary MR analyses employed the inverse-variance weighted (IVW) method, supplemented by the robust adjusted profile score (RAPS), MR-PRESSO, weighted median, MR-Egger regression, and IVW radial analyses. Bidirectional and multivariable MR were also performed. Sensitivity analyses for heterogeneity and horizontal pleiotropy were conducted to ensure robustness. No micronutrient passed FDR < 0.05, but three achieved suggestive evidence. Higher genetically predicted retinol and magnesium were associated with lower RLS risk (retinol OR 0.46, 95% CI 0.25-0.86; magnesium OR 0.62, 0.39-0.98). After excluding two pleiotropic SNPs, higher folate was associated with higher risk (OR 1.48, 1.10-2.00). Reverse MR showed no effect of RLS liability on these nutrients. In multivariable MR, folate remained positively associated (OR 2.88, 1.07-7.77) and magnesium inversely associated (OR 0.38, 0.15-0.98); the retinol signal weakened (OR 0.55, 0.30-1.01). No other micronutrient demonstrated a causal link. Sensitivity tests showed no material heterogeneity or directional pleiotropy. Genetic evidence supports folate excess and magnesium insufficiency as independent, potentially modifiable contributors to RLS, whereas any protective effect of retinol appears sensitive to joint modeling. These findings warrant replication in non-European cohorts and mechanistic studies addressing brain-specific micronutrient regulation.