We explore the connection between different classes of active galactic nuclei (AGNs) and the evolution of their host galaxies, by deriving host galaxy properties, clustering, and Eddington ratios of AGNs selected in the radio, X-ray, and infrared. We study a sample of 585 AGNs at 0.25 < z < 0.8 using redshifts from the AGN and Galaxy Evolution Survey (AGES) and data in the radio (WSRT 1.4 GHz), X-rays (Chandra XBootes), and mid-IR (IRAC Shallow Survey). The radio, X-ray, and IR AGN samples show modest overlap, indicating that to the flux limits of the survey, they represent largely distinct classes of AGNs. We derive host galaxy colors and luminosities, as well as Eddington ratios (lambda), for obscured or optically faint AGNs. We also measure the two-point cross-correlation between AGNs and galaxies on scales of 0.3-10 h^-1 Mpc, and derive typical dark matter halo masses. We find that: (1) radio AGNs are mainly found in luminous red galaxies, are strongly clustered (with M_halo ~ 3x10^13 h^-1 M_sun), and have very low lambda <~ 10^-3; (2) X-ray-selected AGNs are preferentially found in galaxies in the "green valley" of color-magnitude space and are clustered similarly to typical AGES galaxies (M_halo ~ 10^13 h^-1 M_sun), with 10^-3 <~ lambda <~ 1; (3) IR AGNs reside in slightly bluer, less luminous galaxies than X-ray AGNs, are weakly clustered (M_halo <~ 10^12 h^-1 M_sun), and have lambda > 10^-2. We interpret these results in terms of a simple model of AGN and galaxy evolution, whereby a "quasar" phase and the growth of the stellar bulge occurs when a galaxy's dark matter halo reaches a critical mass between ~10^12 and 10^13 M_sun. Subsequently, star formation ceases and AGN accretion shifts from radiatively efficient (optical- and IR- bright) to radiatively inefficient (optically-faint, radio-bright) modes.