We introduce methods which allow observed galaxy clustering to be used together with observed luminosity or stellar mass functions to constrain the physics of galaxy formation. We show how the projected two-point correlation function of galaxies in a large semi-analytic simulation can be estimated to better than ~10% using only a very small subsample of the subhalo merger trees. This allows measured correlations to be used as constraints in a Monte Carlo Markov Chain exploration of the astrophysical and cosmological parameter space. An important part of our scheme is an analytic profile which captures the simulated satellite distribution extremely well out to several halo virial radii. This is essential to reproduce the correlation properties of the full simulation at intermediate separations. As a first application, we use low-redshift clustering and abundance measurements to constrain a recent version of the Munich semi-analytic model. The preferred values of most parameters are consistent with those found previously, with significantly improved constraints and somewhat shifted "best" values for parameters that primarily affect spatial distributions. Our methods allow multi-epoch data on galaxy clustering and abundance to be used as joint constraints on galaxy formation. This may lead to significant constraints on cosmological parameters even after marginalising over galaxy formation physics.