Purpose To introduce a noninvasive eye fixation and computer-aided eye monitoring system for linear accelerator-based stereotactic radiotherapy for uveal melanoma. Methods and materials At the Department of Radiotherapy and Radiobiology, University of Vienna, stereotactic radiotherapy is offered to patients with uveal melanoma considered unsuitable for 106Ru brachytherapy or local resection. For the present feasibility study, 8 patients were carefully selected according to their ability to fixate a small light source with the diseased eye and whether they had a rather small head to meet the limited geometric space available. A polymethyl methacrylate tube was attached to a stereotactic mask system in craniocaudal orientation supporting a 45° mirror, which was placed in front of the diseased eye. At the other end of the tube, the patient was given a small fixation light, and a small camera was positioned beneath, which was shielded for use during MRI. A computer interface calculated and visualized the spatial difference of the actual and a given reference pupil position, which was defined before CT scanning, during the MRI sequences, and during treatment delivery at the linear accelerator. Results The described system can be attached to a conventional stereotactic mask system with minor modifications. Because of the large distance between the eye and the fixation light, the optical fixation system was well tolerated by all patients, and a stable position of the eye was obtained. The camera system can be used during CT and MRI without interference. Absorption of the 6-MV photon beam by the mirror and the polymethyl methacrylate tube was negligible. The computer interface designed to determine the pupil position uses an image-processing algorithm that correlates a template of the reference image with the actual image of the eye. Provided sufficient illumination of the pupil, the correlation function showed a pronounced minimum at the reference position. The precision of the algorithm was tested by phantom measurements. For a given 1 mm or 2 mm displacement, the interface reported a mean shift of 0.96 ± 0.18 mm or 2.07 ± 0.11 mm, respectively. Conclusion The results of this study demonstrated the feasibility of a new optical fixation system for linear accelerator-based stereotaxis. The artifact-free application of the camera system during image acquisition and irradiation and the use of the computer interface, which automatically monitored eye movements with submillimeter precision, provided large improvements compared with existing techniques. Given well-defined interruption criteria and accelerated image processing, the described system has a high potential to perform automatically gated treatment beam delivery in the near future.