Robot-assisted vascular interventional surgery (RVIS) is an emerging technology for the treatment of vascular diseases. It has obvious advantages over traditional manual operation, such as increased accuracy, reduced fatigue, and reduced tremor. However, current research suggests that natural human–robot interaction in RVIS is still a challenge that needs to be addressed. In this article, we developed a novel robotic platform that realized magnetorheological (MR) fluids-based haptic feedback to improve the interventionist’s tactile presence. In addition, we proposed a force sensing method to accurately detect the real-time force of the flexible instrument and a collaborative operation method of the guidewire and the catheter to assist the flexible instruments in selecting the target blood vessel branch and reduce the operation difficulty of the interventionist. To verify the developed robotic platform and the proposed methods, we conducted the performance evaluation experiments in a blood vessel model and an endo vascular evaluator. The results indicated that the developed robotic platform and the proposed methods have great potential to improve the natural human–robot interaction in RVIS and guarantee safety.