Accurate, sensitive and selective detection of metabolic biomarkers in biofluids are of vital significance for health self-monitoring and chronic disease prevention. Here, for the first time, a smart dual-responsive nanozyme sensor (DNS) was developed for simultaneous analysis of glucose and caffeine utilizing stimuli-responsive yolk-shell gold nanoparticles (GNPs)-embedded MIL-53 (Al) (GNPs@MIL-53) structures. After the introduction of glucose, GNPs@MIL-53 displays excellent glucose oxidase (GOx)-like activity to induce the conversion of glucose to gluconic acid and H2O2. H2O2 can oxidize 3,3',5,5'-tetramethylbenzidine (TMB) with the generation a bright-blue color, enabling in-field visualization and surface enhanced Raman scattering (SERS) detection of glucose. Upon the addition of caffeine, 2-aminoterephthalic acid modified MIL-53 can react with the caffeine to form intermolecular hydrogen-bonded complexes, leading to strong cyan fluorescence and significant Raman enhancements. The DNS with multi-channel signal outputs can simultaneously determine glucose and caffeine at concentrations of as low as 3 × 10-8 M and 1.2 × 10-11 M, respectively. Importantly, the DNS-based analytical system not only enables visual discrimination and accurate assay of glucose and caffeine in biofluids, but also exhibits negligible cross-interference between glucose and caffeine determination. The combined characteristics of high selectivity, enhanced accuracy and superior quantitative performance make our platform suitable for the point-of-care monitoring of chronic-disease-related metabolic biomarkers.