The D4000 spectral break index is one of the most important features in the visible spectrum, as it is a proxy for stellar ages and is also used in galaxy classification. However, its direct measurement has always been reserved to spectroscopic observations. In this work, we present a general method to directly measure the D4000 spectral break index with narrow-band photometry: this method has been validated using realistic simulations, and then evaluated with PAUS NBs, cross-matched with VIPERS spectra (𝑖 AB < 22.5, 0.562 < 𝑧 < 0.967). For comparison, we also determine the D4000 in this sample with the SED-fitting code CIGALE (for both PAUS NBs and broad-band data from CFHTLS). The direct D4000 measurement has significantly lower SNR than the CIGALE D4000, however, we find that for 𝑖 AB < 21 (two magnitudes above PAUS completeness limit) all direct D4000 measurements have 𝑆𝑁 𝑅 > 3. Moreover, CIGALE underestimates the error by >50%, while the direct D4000 has proper error estimation within <10%. We study the D4000-stellar mass relationship for red and blue galaxies, as well as the D4000-SFR dependence: all methods show agreements with VIPERS within 1𝜎, and the D4000-mass relation is especially well reproduced for blue galaxies. We also evaluate how a D4000 cut classifies galaxies into red/blue compared to machine learning: the CIGALE D4000 with PAUS NBs has the best results while providing realistic cut values. We conclude that the direct measurement of D4000 with narrow-band photometry is a promising tool that allows individual measurements for objects measured with 𝑆𝑁 𝑅 > 3, or averages by stacking, with results compatible with spectroscopy.