Amyotrophic lateral sclerosis (ALS) is characterized by the absence of reliable diagnostic biomarkers. The aim of the study was to (i) devise an untargeted metabolomics methodology that reliably compares cerebrospinal fluid (CSF) from ALS patients and controls by liquid chromatography coupled to high-resolution mass spectrometry (LC-HRMS); (ii) ascertain a metabolic signature of ALS by use of the LC-HRMS platform; (iii) identify metabolites for use as diagnostic or pathophysiologic markers. We developed a method to analyze CSF components by UPLC coupled with a Q-Exactive mass spectrometer that uses electrospray ionization. Metabolomic profiles were created from the CSF obtained at diagnosis from ALS patients and patients with other neurological conditions. We performed multivariate analyses (OPLS-DA) and univariate analyses to assess the contribution of individual metabolites as well as compounds identified in other studies. Sixty-six CSF samples from ALS patients and 128 from controls were analyzed. Metabolome analysis correctly predicted the diagnosis of ALS in more than 80% of cases. OPLS-DA identified four features that discriminated diagnostic group (p < 0.004). Our data demonstrate that untargeted metabolomics with LC-HRMS is a robust procedure to generate a specific metabolic profile for ALS from CSF and could be an important aid to the development of biomarkers for the disease.