The most common gynecological malignancy in developed countries is endometrial cancer (EC). Early diagnosis is crucial to improve the disease prognosis, which strongly relies on the tumor stage. Metabolomics has emerged as a promising noninvasive test for various diseases, including EC. However, no metabolic marker has been authorized for routine use. Various sample types may be used to extract metabolites for EC detection, including endometrial tissue, blood samples, and urine. While sampling the endometrial cavity directly has great potential for yielding reliable EC-related biomarkers, this invasive method limits the clinical utility. Metabolites in blood are more easily accessed, but a low yield reduces the diagnostic potential, particularly in early-stage tumors. Another possibility is extracting metabolites from easily accessed media such as tampons, vaginal swabs, and cervicovaginal sampling which is minimally invasive. Amino acid, lipid, and hormonal metabolites have shown potential as biomarkers for EC detection and prognosis as well as monitoring for EC relapse. To validate these metabolites as clinically reliable markers, and their role in EC tumorigenesis, further studies are needed. The use of molecules detected through metabolomics has great clinical potential, especially if the source tissue is collected with minimal patient invasion. The Cancer Genome Atlas (TCGA) has named specific EC molecular subtypes that offer the better probability for prognostic discrimination versus traditional EC histological subtypes, future studies address these metabolites. Advances in artificial intelligence and machine learning techniques will surely assist and enable the production of a clinically reliable EC detection metabolomic biomarker panel.