Abstract Objectives The stratification of individuals suffering from acute and post-acute SARS-CoV-2 infection remains a critical challenge. Notably, biomarkers able to specifically monitor viral progression, providing details about patient clinical status, are still not available. Herein, quantitative metabolomics is progressively recognized as a useful tool to describe the consequences of virus-host interactions considering also clinical metadata. Methods The present study characterized the urinary metabolic profile of 243 infected individuals by quantitative nuclear magnetic resonance (NMR) spectroscopy and liquid chromatography mass spectrometry (LC–MS). Results were compared with a historical cohort of noninfected subjects. Moreover, we assessed the concentration of recently identified antiviral nucleosides and their association with other metabolites and clinical data. Results Urinary metabolomics can stratify patients into classes of disease severity, with a discrimination ability comparable to that of clinical biomarkers. Kynurenines showed the highest fold change in clinically-deteriorated patients and higher-risk subjects. Unique metabolite clusters were also generated based on age, sex, and body mass index (BMI). Changes in the concentration of antiviral nucleosides were associated with either other metabolites or clinical variables. Increased kynurenines and reduced trigonelline excretion indicated a disrupted nicotinamide adenine nucleotide (NAD⁺) and sirtuin 1 (SIRT1) pathway. Conclusions Our results confirm the potential of urinary metabolomics for noninvasive diagnostic/prognostic screening and show that the antiviral nucleosides could represent novel biomarkers linking viral load, immune response, and metabolism. Moreover, we established for the first time a casual link between kynurenine accumulation and deranged NAD⁺/SIRT1, offering a novel mechanism through which SARS-CoV-2 manipulates host physiology.