Current industry practice to connect composite structural panels still relies on the extensive use of bolted joints. However, due to their anisotropic nature, composite materials joining techniques represent a true design challenge, as their behaviour can change with, e.g. layup, material, geometry and environmental conditions. For this reason, the behaviour of composite bolted joints has been the subject of research studies for many years. Particular focus has been given to in-plane and out-of-plane performance of mechanically fastened composites. However, in real-life applications, e.g. for L-junctions and single lap joints of thin laminates, mechanically fastened joints are subject not only to pure bearing or pure pull-through, but have to sustain a combination of both loading scenarios [1,2]. Here, a novel test rig for determining the failure envelope of composite fastened joints under combined bearing and pull-through loading conditions is designed, tested and validated. The proposed device is compatible with a standard testing machine and captures the critical interactions between the two loading mechanisms for pull-through to bearing ratios ranging from 0.2 up to 0.8, allowing the characterization of the behaviour of composite joints under realistic loading scenarios. The multiaxial failure envelope for a carbon-fibre-based composite laminate is obtained experimentally and the performed tests are shown to fall within the limit values obtained for the pure bearing and pure pull-through loading conditions, obtained following well-established ASTM standards [3,4]. High-fidelity finite element simulations [5] are used to support the validation of the test methodology: the ability of the fixture to maintain a constant mixed load ratio during the whole test was numerically confirmed and a detailed analysis of the failure modes provided further insights into the main degradation mechanisms occurring during the experiments. A very good agreement between the numerical and experimental results is obtained for the entire failure envelope, with the numerical predictions falling between +-10% relative errors, further validating the constitutive modelling predictive capabilities.