Background Endometrial cancer (EC) is a common malignant tumor in women with increasing mortality. The prognosis of EC is highly heterogeneous which needs more effective biomarkers for clinical decision. Here, we reported the effect of autophagy-related genes (ARGs) on the prognosis of EC. Methods The expression data of EC tissues and adjacent non-tumor samples were available from the TCGA dataset and 232 autophagy-related genes were from The Human Autophagy Database. A prognostic ARGs risk model was further constructed by using LASSO-Cox regression, and its prognostic and predictive value were evaluated by nomogram. Further functional analysis was conducted to reveal a significant signaling pathway. Results A total of 45 differentially expressed ARGs were obtained, including 18 upregulated and 27 downregulated genes. Eleven ARGs (BID, CAPN2, CDKN2A, DLC1, GRID2, IFNG, MYC, NRG3, P4HB, PTK6, and TP73) were finally selected to build ARGs risk. This signature could well distinguish between the high- and low-risk patients (survival analysis: P  = 1.18E-10; AUC: 0.733 at 1 year, 0.795 at 3 years, and 0.823 at 5 years). Furthermore, a nomogram was plotting to predict the possibility of overall survival and suggested good value for clinical utility. Conclusion We established an eleven-ARG signature, which was probably effective in the prognostic prediction of patients with EC.