Schizophrenia (SCZ) and bipolar disorder (BD) are serious psychiatric disorders and share several characteristics and the diagnosis yet is mainly clinical. The sooner they are identified, diagnosed and treated, the better the clinical prognosis. Therefore, the development of sensitive and accurate biomarkers is highly required. Lipids play an increasingly recognized role in the neuronal function and plasticity of the brain. Glycerophospholipids and molecules-like comprise 60% of the non-aqueous portion of the brain and in an even greater proportion of the dendrites and synapses. Other metabolites directly influence its functioning and remodeling, such as acylcarnitines, sphingolipids, cholesterol and other lipids. Since lipid metabolism is altered differently in neuropsychiatric diseases, alterations in the lipid profile of the membrane can allow a discrimination between subjects in first-episode psychosis). Thus, our aim was to determine plasma levels of metabolites of subjects in FEP and controls and find cutoff values that differentiate each group. Plasma samples were analyzed for 55 drug-naïve patients (28 SCZ and 27 BD) and 30 controls. Determining the lipid profile was performed by mass spectrometry - Flow injection analysis, using AbsoluteIDQ p180® kit (Biocrates Life Sciences). Statistical analyzes were performed using a classification method - Classification And Regression Tree. We observed that the combination of four metabolites are able to differentiate the diagnoses: PCaaC26:0, PCaaC38:4, PCaaC34:3 and C16-OH. The accuracy of the method is 87,1%. Discussion: Our results suggest that the levels of some plasma metabolites differentiate subjects with SCZ, BD and controls. The levels of these metabolites can be a potential biomarker for psychosis, as well as a diagnostic marker for SCZ and BD. The findings from this study require further validation in BD and SCZ subjects, but suggest that the metabolome is a good tool to understand the pathophysiology of these disorders and presents potential diagnostic biomarkers.