Abstract: Background　Neuroendoscopic port surgery is the mainstream approach for endoscopic treatment of hypertensive intracerebral hemorrhage, and its application is gradually expanding. Precise, stable scope manipulation by an assistant is critical but challenging due to the steep learning curve of manual holding. Hence, a novel automated holding system featuring ease of use, rapid response, and cost efficiency is urgently warranted. Objective To assess the feasibility and imaging performance of a 3D vision - guided collaborative robot (CoBot) serving as a neuroendoscope holder assisting endoscope reposition in an in vitro port model. Methods　A 0° training neuroendoscope was mounted on a 6-degree-of-freedom robotic arm equipped with a 3D vision camera to recognize a marker on the endoscopic port sheath and generate a real-time repositioning trajectory. In Test 1, 30 consecutive repositioning cycles were performed to quantify workflow success rate, recognition time, and motion time. In Test 2, the robot repeated the workflow for 30 cycles and captured endoscopic images at a preset working distance. In Test 3, experienced neuroendoscopic surgeons from our hospital manually executed the same workflow under identical port and endoscope settings. Blinded experts rated the images from Tests 2 and 3 on a 5-point Likert scale for clarity, accuracy, and stability. Results In the feasibility test, the robot completed all 30 cycles without collision (100% success rate), with a mean recognition time of (2.37±0.05) s and a mean motion time of (2.98±0.35) s. For image quality scores (presented as median and interquartile range), the robotic group demonstrated significantly higher median scores than the manual group across all three dimensions: clarity (5.00 5.00 - 5.00 vs 4.00 3.00 - 4.00), accuracy (5.00 4.00 - 5.00 vs 3.00 3.00 - 4.00), and stability (4.00 4.00 - 5.00 vs 3.00 3.00 - 4.00) (all P＜0.001). Conclusion Under in vitro conditions, our CoBot system achieves stable and reproducible endoscopic repositioning with superior image quality compared to manual holding, providing preliminary evidence for further clinical translation.