Breath biopsy is emerging as a rapid and non-invasive diagnostic tool that links exhaled chemical signatures with specific medical conditions. Despite its potential, clinical translation remains limited by the challenge of reliably detecting endogenous, disease-specific biomarkers in breath. Synthetic biomarkers represent an emerging paradigm for precision diagnostics such that they amplify activity-based biochemical signals associated with disease fingerprints. However, their adaptation to breath biopsy has been constrained by the limited availability of orthogonal volatile reporters that are detectable in exhaled breath. Here, we engineer multiplexed breath biomarkers that couple aberrant protease activities to exogenous volatile reporters. We designed novel intramolecular reactions that leverage protease-mediated aminolysis, enabling the sensing of a broad spectrum of proteases, and that each release a unique reporter in breath. This approach was validated in a mouse model of influenza to establish baseline sensitivity and specificity in a controlled inflammatory setting, and subsequently applied to diagnose lung cancer using an autochthonous Alk -mutant model. We show that combining multiplexed reporter signals with machine learning algorithms enables tumor progression tracking, treatment response monitoring, and detection of relapse after 30 minutes. Our multiplexed breath biopsy platform highlights a promising avenue for rapid, point-of-care diagnostics across diverse disease states.