This paper presents a global paired Argo dataset documenting the upper-ocean temperature and salinity responses to tropical cyclones (TCs) during 2000–2024. The dataset was built by combining the International Best Track Archive for Climate Stewardship (IBTrACS, v04r01) with the Global Ocean Argo Scatter Dataset (Version 3.0) provided by the China Argo Real-time Data Center. On top of the original Argo quality-control flags, we applied a strict secondary screening that retained only profiles with good quality flags for temperature, salinity, and pressure at every vertical level and with a maximum sampling depth of at least 1900 m. Each TC location was first matched to candidate floats within an inclusive 1000 km search envelope, with the archived storm-relative coordinates allowing users to apply narrower distance thresholds for specific analyses. The two profiles in each pair were further required to lie within 50 km of each other to suppress horizontal mesoscale aliasing. All retained profiles were linearly interpolated onto 56 standard depth levels from the surface to 2000 m and rotated into a storm-relative Cartesian frame in which the TC center is the origin and the translation direction is aligned with the positive Y axis. The final product contains 3164876 high-quality paired records—2747731 in the Northern Hemisphere and 417145 in the Southern Hemisphere—archived as six self-describing netCDF files separated by hemisphere and time period. Spatial, vertical, and temporal composites confirm that the dataset reproduces the canonical upper-ocean response to TCs, including the hemispheric mirror asymmetry of cross-track surface cooling, the concentration of the strongest thermal anomalies within the upper ~200 m, and a peak sea surface cooling about 2.5 d after passage followed by a 1–2 week recovery. The salinity response is weaker and less spatially coherent, reflecting the competing effects of rainfall freshening, vertical mixing, and upwelling. By providing background-field profiles, response-field profiles, and their precomputed differences in a single analysis-ready record, this dataset substantially reduces the preprocessing burden for studies of TC–ocean interaction, upper-ocean heat and salt budgets, mixed-layer and near-inertial processes, and the evaluation of coupled ocean–atmosphere models. The dataset and associated data-generation code are available at <a href="https://doi.org/10.6084/m9.figshare.32820563" target="_blank" rel="noopener">https://doi.org/10.6084/m9.figshare.32820563</a> (Wang et al., 2026b).