The mechanisms that drive the development of gravitational instabilities, leading to the formation of the large-scale structure of the universe, is not yet understood in its full details. Yet, with the arrival of a new generation of projects of observational cosmology, such as the LSST and EUCLID, that aim at measuring dark energy properties from large-scale structure observations, it is now necessary to characterize its properties with a large precision and in a controlled manner. N-body simulations bring answers to those issues but only for a very limited number of models and for a limited range of cosmological parameters. The scientific cases of the project aforementioned rely however heavily on our ability to make such predictions for large classes of models. It is therefore necessary to sharpen our theoretical knowledge on the growth of gravitational structures. This project aims at developing tools for predicting and computing cosmic density spectra and bispectra (three-point correlation function) for a large set of cosmological models that include non-standard effects such as massive neutrinos or clustering dark energy. More precisely we wish to build theoretical tools for predicting those quantities analytically and with a controlled precision in the quasilinear regime - therefore in a regime that defines the validity of the linear regime and extends it - and develop robust and fast numerical codes for computing a set of well defined observables such as those related to cosmic shear observations, redshift space density field, etc. We also wish to construct more phenomenological models that explore the relationships between the density field (and its various components) and the halo density. The approaches we favor in this project make use of computation techniques that have been recently put forward in which re-summations of large classes of diagrams can be taken into account. These approaches explicitly, or implicitly, take advantage of the so-called eikonal approximation. Those approaches allow to develop perturbation theory calculations in a controlled way and for a large class of observables such as spectra, bispectra etc. Our project aims at writing and releasing packages - in fortran to make its portability to different systems easier - for the fast computation of perturbation theory spectra and bispectra beyond linear theory. More precisely, we wish to develop codes that compute spectra up to 2 loops (NNLO) and bispectra up to 1 loop (NLO).