Abstract A newly constructed multidirectional cyclic direct simple shear (mcDSS) device with unique capabilities is introduced. This mcDSS apparatus, called the Illinois mcDSS (or I-mcDSS) device, for the first time brings together the following capabilities: (1) servo-hydraulic control that can apply stress- or strain-based monotonic, cyclic (e.g., sinusoidal, saw tooth, and square), and high-frequency broadband loads at realistic earthquake loading rates, improving over previous devices with pneumatic control; (2) unidirectional and bidirectional loading; (3) bender elements to measure (S-wave) velocity; (4) a cell for applying consolidation stresses different from at-rest values as well as back-pressure saturation; and (5) a multidirectional load cell on top of the specimen to minimize the effect of compliance and component friction on load measurements. The I-mcDSS device tests cylindrical specimens confined using either a wire-reinforced membrane or stacked rings. Experiments conducted on a uniformly graded Ottawa sand are presented to illustrate each key I-mcDSS feature. Test repeatability is demonstrated for monotonic and bidirectional cyclic tests. Drained or constant volume, K0-consolidated, strain-controlled monotonic, unidirectional cyclic, and bidirectional (circular, figure-8, and broadband) cyclic tests on dry specimens yielded shear stress–shear strain relations, peak effective-stress friction angles (ϕ′peak-DSS), and volumetric strains, or excess pore water pressures, consistent with the literature. However, depending on load cell and displacement transducer locations, device compliance was observed to affect shear stress–shear strain response at shear strains less than 1 %. When compared with triaxial compression (TC) peak effective-stress friction angles (ϕ′peak-TC), ϕ′peak-DSS was about 5° smaller than ϕ′peak-TC if the horizontal plane is considered the failure plane, whereas ϕ′peak-DSS ≈ ϕ′peak-TC if the horizontal plane is considered the plane of maximum shear stress. Lastly, measured S-wave velocities at varying confinements are consistent with published correlations.