In the last decades, evidence has brought the standard cosmological model to include a significant dark sector with the elusive dark energy and dark matter (DM). Their mainly unknown properties constitute the most pressing questions in cosmology. As the main contributor to the mass content of the Universe, the DM can be observed through its gravitational interaction with normal matter (denoted baryonic in contrast). This interaction is particularly visible in DM dominated environments as galaxy clusters. Galaxy clusters are the largest structures bound by gravity in the Universe. Such massive objects bend the light from background objects, allowing us to study their mass contents with the gravitational lensing effect. In that regard, they are offering us a unique natural laboratory to study the DM properties. However, gravitational lensing alone can only probe the total mass along the line of sight. To go beyond this limitation, we need to combine multiple mass probes within lens mass models to fully disentangle baryons and DM. The work presented in this manuscript aimed at tackling that limitation and developing a self-consistent method that relies on different mass probes to complete the lensing observations and specifically constrains the baryonic content of clusters. The first work presented in this manuscript is dedicated to improving the mass distribution recovered by lensing. A limitation of lensing features is their scarcity, which limits the complexity of mass models. However, with the increasing abilities of observing facilities, the number of lensing features has reached the limit of current model parametrisation. We proposed a perturbative approach to mitigate that issue and go beyond the elliptical shapes provided by analytical mass profiles. A perturbative surface of B-splines is added onto the lensing potential to slightly redistribute the matter in its field of effect. We demonstrate its ability in a realistic simulation of galaxy clusters to improve the reconstruction of the lensing effect and the accuracy of the 1D and 2D mass distribution. Its application on the cluster MACS J1206.2-0847 shows the capabilities of such surface to mimic the effect of "dark haloes". We will then develop the multi-probe mass method by focusing on the intra-cluster gas. It is heated to high temperatures through the gravitational interaction and is in the form of a plasma. In those conditions, the gas makes clusters appear as diffuse X-ray sources, which we can study through space-based X-ray observatories. The gas mass distribution can be mapped thanks to the X-ray surface brightness and spectra. We will show how to include a gas mass component in lensing models in a self-consistent approach where X-ray and lensing constraints are used jointly to recover the total and gas mass simultaneously. Finally, we will be interested in the stellar content of galaxies and intra-cluster stars. The intra-cluster stars appear as a faint diffuse cluster-scale light emission (i.e. Intra-Cluster light). Baryons in both galaxies and the intra-cluster stars can be constrained through their stellar populations and stellar kinematics. We demonstrate that thanks to these stellar observations we are able to fully disentangle the DM from the baryons on cluster- and galaxy-scale for the first time. The constraints on mass profile are strengthened, which reduces the uncertainties on the total or DM mass profile to less than 3%.