Objectives Observational studies have found potential associations between dietary intake and multiple sclerosis (MS). However, these associations are inconsistent, and the causal relationship remains unclear. In this study, we aim to examine the causal relationship between genetically predicted dietary composition and the risk and severity of MS using two-sample Mendelian randomization. Method Genetic instruments for 30 different dietary compositions were extracted from large-scale genome-wide association studies (GWAS), mainly from the UK Biobank dataset. The GWAS data for MS risk and severity were obtained from the International Multiple Sclerosis Genetics Consortium. The primary analysis employed either the inverse variance weighted method or the Wald ratio method to evaluate the causal association. Several sensitivity analyses were also performed. Results Genetically predicted higher pork intake was causally associated with an increased risk of MS (odds ratio [OR] = 6.76; p = 0.005), while genetically driven higher cereal intake (OR = 0.43, p = 0.016), vitamin C supplement (OR < 0.01; p = 4.34 x 10-5), folic acid supplement (OR < 0.01; p = 4.91 x 10-71), and fish oil supplement (OR = 0.04; p = 0.017) were causally linked to a decreased risk of MS. In addition, genetically predicted higher alcoholic intake (OR = 1.17; p = 0.041) was causally associated with an increase in MS severity, while folic acid supplement (OR < 0.01; p = 0.015) was causally linked to a decrease in MS severity. Interpretation This study found that increased consumption of cereal, vitamin C, folic acid, and fish oil, coupled with reduced pork and alcohol intake, may potentially decrease the risk and severity of MS. These findings inform the development of dietary-based strategies for MS prevention and treatment.