Résumé : The AGN bolometric correction is a key element to understand BH demographics and compute accurate BH accretion histories from AGN luminosities. However, current estimates still differ from each other by up to a factor of two to three, and rely on extrapolations at the lowest and highest luminosities. Here we revisit this fundamental issue presenting general hard X-ray ($K_{X}$) and optical ($K_{O}$) bolometric corrections, computed combining several AGN samples spanning the widest (about 7 dex) luminosity range ever used for this kind of studies. We analysed a total of $\sim 1000$ type 1 and type 2 AGN for which a dedicated SED-fitting has been carried out. We provide a bolometric correction separately for type 1 and type 2 AGN; the two bolometric corrections results to be in agreement in the overlapping luminosity range and therefore, for the first time, a universal bolometric correction for the whole AGN sample (both type 1 and type 2) has been computed. We found that $K_{X}$ is fairly constant at $log(L_{BOL}/L_{\odot}) < 11$, while it increases up to about one order of magnitude at $log(L_{BOL}/L_{\odot}) \sim 14.5$. A similar increasing trend has been observed when its dependence on either the Eddington ratio or the BH mass is considered, while no dependence on redshift up to $z\sim3.5$ has been found. On the contrary, the optical bolometric correction appears to be fairly constant (i.e. $K_{O} \sim 5$) whatever is the independent variable. We also verified that our bolometric corrections correctly predict the AGN bolometric luminosity functions. According to this analysis, our bolometric corrections can be applied to the whole AGN population in a wide range of luminosity and redshift.