We formulate a method to study two-body correlations in a condensate of N identical bosons. We use the adiabatic hyperspheric approach and assume a Faddeev like decomposition of the wave function. We derive for a fixed hyperradius an integro-differential equation for the angular eigenvalue and wave function. We discuss properties of the solutions and illustrate with numerical results. The interaction energy is for N ≈ 20 five times smaller than that of the Gross-Pitaevskii equation. PACS number(s): 03.75.Fi, 05.30.Jp Introduction. Few-body correlations often express the distinguishing characteristic features of an N -body system [1]. Two-body correlations are not only the simplest but in most cases also the most important. Higher order correlations have a tendency to either strongly confine spatially or correlate into clusters of particles effectively reducing the correlations to lower order. Exceptions are the three-body correlations decisive for the stability of Borromean systems known for dripline nuclei [2].Non-correlated mean-field computations of nuclei with the free two-body nucleon-nucleon interaction produce disastrously wrong results. Two-body correlations compensating for the short range hard core repulsion are absolutely necessary [3]. For atoms the effective interaction is also strongly repulsive at shorter distances. Furthermore for an atomic Bose condensate a short-range twobody attraction produces diatomic recombination and thereby the atoms decay out of the condensate [4]. For both nuclei and molecules correlations are decisive. For nuclei various methods have been designed to deal with this problem, i.e. Jastrow theory, Bruckner theory, effective interactions in model spaces [5].For Bose condensates the Gross-Pitaevskii (noncorrelated) mean-field equation has been the starting point since the first observation of the condensate in 1995 [6]. The wave function does not include any correlations and the assumed repulsive δ-interaction has the immediate consequence that the short range behavior cannot be described even qualitatively correct. Repulsive zero range potentials are not physically meaningful [7] although sometimes useful in selected Hilbert spaces. Attractive δ-interactions in three dimensions lead to divergencies demanding renormalization [8,9] or a change of boundary conditions [10]. When two-body bound states appear even dimer condensates may be possible [11].To include correlations we must necessarily go beyond the mean-field Hartree-Fock-Boguliubov approximation. Then finite range potentials with realistic features can as well be used as the starting point in the theoretical formulation. The other crucial ingredient is the degrees of freedom or, equivalently, the Hilbert space. An interesting formulation was recently introduced in terms of generalized hyperspherical coordinates and an adiabatic expansion with the hyperradius as the adiabatic