Recently, a commercial system capable of x-ray image guided patient positioning and respiratory gated delivery has become available. Here we describe the operational principles of this system and investigate its geometric targeting accuracy under controlled conditions. The system tracks breathing via infrared (IR) detection of reflective markers located on the patient's abdomen. Localization kilovoltage (kV) x-rays are triggered from within the gated delivery window portion of the breathing trace and after positioning, the tumour will cross the linac isocentre during gated delivery. We tested geometric accuracy of this system by localizing and delivering gated fields to a moving phantom. Effects of phantom speed, gating window location, timing errors and phantom rotations on positioning and gating accuracy were investigated. The system delivered gated fields to both a moving and static phantom with equal accuracy. The position of the gating window affects accuracy only to the extent that an asymmetric breathing motion could affect dose distribution within its boundaries. Positioning errors were found to be less then 0.5 +/- 0.2 mm for phantom rotations up to 5 degrees. We found and corrected a synchronization error caused by a faulty x-ray duration setting and detected a 60 +/- 20 ms time delay in our linear accelerator.