The pathogenesis of frozen shoulder (FS) remains unclear, and current research primarily focuses on immune responses. Increasing evidence suggests that immune cells play a significant role in FS development. However, the causal relationship between the two remains poorly understood. Therefore, we aimed to investigate this using Mendelian randomization (MR) analysis. Single nucleotide polymorphisms closely associated with 731 immune phenotypes were obtained from publicly available GWAS datasets as instrumental variables. FS was used as the outcome with a sample size of 451,099 cases. Causal effects were analyzed using the inverse variance-weighted method. We conducted sensitivity tests, including the intercept of the MR-Egger and MR-PRESSO analyses. The presence of heterogeneity was evaluated using Cochran Q test. We identified potential causal relationships in terms of increased risk for FS with 5 immune phenotypes: CD25++ CD45RA+ CD4 not regulatory T cell %CD4+ T cells (odds ratio [OR] = 1.0273, 95% confidence interval [CI]: 1.0093–1.0457, P = .0028), CD25++ CD45RA+ CD4 not regulatory T cell %T cell (OR = 1.0240, 95% CI: 1.0057–1.0427, P = .0098), CD127 on CD28+ CD4+ T cells (OR = 1.0398, 95% CI: 1.0121–1.0682, P = .0046), CD4 on human leukocyte antigen DR+ CD4+ T cells (OR = 1.0795, 95% CI: 1.0316–1.2195, P = .0009), and human leukocyte antigen DR on CD14− CD16+ monocytes (OR = 1.0533, 95% CI: 1.0136–1.0945, P = .0081). Few significant heterogeneities or horizontal pleiotropies were observed. Through MR analysis, we identified distinct 5 types of immune cells that were positively correlated with the occurrence and development of FS, providing guidance for clinical intervention in FS.