We combine optical, X-ray (Einstein and ROSAT) and infrared (IRAS) data to investigate the evolution and spatial distribution of AGN with particular emphasis on the implications for the origin of the diffuse X-ray background. First, we derive the IRAS Seyfert luminosity function to test the continuity of properties between the Seyfert and the QSO population. The QSO luminosity function evolved back to z ~ 0, agrees well with the Seyfert luminosity function. In particular, the similarity of the faint parts of the two luminosity functions, suggests that the optical luminosity function is not severely affected by incompleteness due to reddening. We analyze the clustering properties of the IRAS Seyfert sample in order to probe the AGN clustering evolution. We detect clear clustering (5σ) at scales < 20 h(^-1). Comparing the Seyfert with the QSO clustering results at higher redshifts we find that a comoving model for AGN clustering evolution, where AGN clusters are expanding with the Hubble flow, is probably favoured by the data. Using new faint CCD observations we recalibrate the photometry of the Durham UVX catalogue of Boyle et al. (1990). We show that the luminosity function 'knee' feature claimed by Boyle et al. is not an artefact of photometric errors at faint magnitudes. We evaluate the contribution of Active Galactic Nuclei (AGN) to the X-ray background using this optical luminosity function and evolve it according to Pure Luminosity Evolution models. We estimate that AGN produce around half of the X-ray background at 2 keV. This contribution is consistent with the small scale (arcmin) fluctuations of the X-ray background for both the stable and comoving model of clustering evolution. If a large number of low luminosity AGN with high intrinsic absorption is missed by the optical surveys, AGN could produce all the 2 keV intensity. Although the uncertainty in the estimate of the AGN contribution is high, this work demonstrates, at least, that Pure Luminosity Evolution models are consistent with both the X-ray background intensity and anisotropy constraints. A recent deep ROSAT observation yields a high surface density of X-ray sources (> 100 deg(^-2). Spectroscopic follow up observations show that most of these sources are QSOs. The identified QSOs contribute ~ 30% at 1 keV and therefore this is the lower limit of the AGN contribution to the X-ray background. No other class of sources contributing substantially to the X-ray background has been yet identified.