This paper presents the scientific case for a next generation adaptive optics instrument at the VLT, temporarily named “Planet Finder”, that is aimed at detecting and characterizing extrasolar planets through the direct analysis of their emitted photons in the visible and at near-IR wavelengths. We discuss the observational niche of such an instrument to have first light in 2010, in complement to other planet search methods. To improve the efficiency (and consistency) of the search for planets with the PF, the observations will need to be organized in the form of an extensive survey of hundreds of nearby stars, predicted outputs of which are also described here. This summarizes the study phase of the instrument, conducted by two competitive teams and the recent merging of both studies, regarding the scientific impact of Planet Finder.