Despite their biological importance in many regulatory processes, protein-peptide recognition mechanisms are difficult to study experimentally at the structural level because of the inherent flexibility of peptides and the often transient interactions on which they rely. Complementary methods like biomolecular docking are therefore required. The prediction of the three-dimensional structure of protein-peptide complexes raises unique challenges for computational algorithms, as exemplified by the recent introduction of protein-peptide targets in the blind international experiment CAPRI (Critical Assessment of PRedicted Interactions). Conventional protein-protein docking approaches are often struggling with the high flexibility of peptides whose short sizes impede protocols and scoring functions developed for larger interfaces. On the other side, protein-small ligand docking methods are unable to cope with the larger number of degrees of freedom in peptides compared to small molecules and the typically reduced available information to define the binding site. In this chapter, we describe a protocol to model protein-peptide complexes using the HADDOCK web server, working through a test case to illustrate every steps. The flexibility challenge that peptides represent is dealt with by combining elements of conformational selection and induced fit molecular recognition theories.