Optimising haptic sensitivity in visual-haptic interfaces can have important benefits, particularly in situations where visual information is relatively unreliable (for example in surgical devices, where there is often a lack of normal visual depth cues). Many haptic devices alter the normal relationship between object size and hand opening, and so might be expected to alter haptic sensitivity. To determine how to optimise haptic sensitivity in such situations, we measured haptic size sensitivity as a function of object size (i) during normal grasping, and (ii) using pliers-like tools that changed the gain between hand opening and object size. Haptic stimuli were created using force-feedback robots, and we measured discrimination thresholds using a two-interval forced-choice task. We first showed that when participants directly grasped objects with their hands haptic size sensitivity did not follow Weber’s law, but instead varied non-monotonically with object size/hand opening; thresholds were lowest for object sizes around 45 mm and increased non-linearly for larger and smaller objects. Second, when using the pliers, sensitivity in units of hand opening was unchanged from normal grasping. Thus, thresholds in units of object size were simply the product of the non-monotonic haptic size sensitivity function, above, and the tool gain. Taken together, these results show that there is a unique tool gain that maximises haptic sensitivity to each object size. Our results show how the ‘tool gain’ of haptic devices should be altered to optimise sensitivity to the range of object sizes in a particular scene.