his thesis reports on the magnetic properties of bi- and tri-metallic nanostructures at surfaces. The main goal is to build the smallest nanostructure ferromagnetic at room temperature and the assembly of those structures in high density arrays. To this purpose we have developed an approach consisting in a fine engineering of the nanostructure chemical structure at the atomic level. We grow core-shell nanostructures with atomically sharp interfaces between the different constituents and we investigate the effects produced by these interfaces as a function of their chemistry and structure. This approach is then used to grow organized arrays of bimetallic nanostructures that represent model systems for next generation high density magnetic storage media. The magnetic properties of the different experimental systems were investigated by magneto-optical Kerr effect (MOKE) and they were correlated to the morphology obtained by scanning tunneling microscopy (STM). The first part focuses on the understanding of the effects of atomically sharp interfaces among several elements. Two dimensional cobalt nanostructures have been grown on Pt(111) single crystal surface and subsequently their edges have been decorated by Fe, Pt or Pd. This one-dimensional decoration produces element dependent variations of the magnetic anisotropy energy with strong enhancement for Fe decoration and reduction for Pt and Pd. Furthermore we observe a crystallographic direction dependence on the effect of Pt and Pd decoration since Co/Pt or Co/Pd staking with {111}-orientation strongly enhances the magnetic hardness of the nanostructures. The effect of atomically sharp Co/Fe one-dimensional interface has been compared with direct alloying of the two elements finding the first one to give the highest effect for shell thicknesses smaller than 5 atoms. With the help of fully relativistic ab-initio we were able to reproduce the experimental results and unravel their pure electronic origin. The electronic hybridizations at the interface between the elements gives rise to "hot spots" in the electronic band structure responsible of a strong variation of the magnetocrystalline anisotropy resulting in the enhanced magnetic hardness. By growing Co-core Fe-shell Pd-capped nanostructures we were able to enhance the magnetization thermal stability of nanostructures by almost a factor of 3 with respect single-metal nanostructures of the same size. In the second part of this thesis, the first attempt of growing organized array of bimetallic nanostructures is presented. Taking advantage of the well-known self organized growth of Co nanodots on Au(111)-vicinal surfaces, we were able to grow Co-core Fe-shell nanodots on Au(11,12,12) single crystal. The thermal stability enhancement produced by the atomically sharp Co/Fe interface is higher than the direct alloying of the two elements as in the previous case. We argue that with this method, combined with three-dimensional growth of nanostructures arrays, it will be possible to produce multimetallic nanodots composed by ≈ 3000 atoms with magnetization thermal stability suitable for magnetic storage media. In the last part another model system for magnetic nanostructure superlattices is presented. Fe nanostructures have been grown on the Pd-seeded and unseeded alumina thin film formed by oxidation of a Ni3Al(111) single crystal surface at high temperature. We found that the sample stoichiometry near the surface evolves with the preparation cycles. This leads to the formation of a Ni-rich region, close to the alumina surface, ferromagnetic up to 250 K. By means of X-ray magnetic circular dichroism measurements we were able to estimate the amount of Ni in excess. Fe clusters deposited on the alumina surface are magnetically coupled to the Ni-rich region independently on the presence of Pd. Nevertheless, the system behavior cannot be explained by a simple ferromagnetic or antiferromagnetic coupling. We argue that the Fe nanostructures are coupled to the substrate by an interlayer exchange coupling across the alumina thin film. The coupling constant is expected to change sign depending on the distance between the Fe clusters and the Ni-rich region resulting in a peculiar magnetic behavior of the system.