This paper reports on the commissioning of an Elekta cone‐beam computed tomography (CT) system at one of the first U.S. sites to install a “regular,” off‐the‐shelf Elekta Synergy (Elekta, Stockholm, Sweden) accelerator system. We present the quality assurance (QA) procedure as a guide for other users. The commissioning had six elements: (1) system safety, (2) geometric accuracy (agreement of megavoltage and kilovoltage beam isocenters), (3) image quality, (4) registration and correction accuracy, (5) dose to patient and dosimetric stability, and (6) QA procedures. The system passed the safety tests, and agreement of the isocenters was found to be within 1 mm. Using a precisely moved skull phantom, the reconstruction and alignment algorithm was found to be accurate within 1 mm and 1 degree in each dimension. Of 12 measurement points spanning a 9×9×15‐cm volume in a Rando phantom (The Phantom Laboratory, Salem, NY), the average agreement in the x, y, and z coordinates was 0.10 mm, −0.12 mm, and 0.22 mm [standard deviations (SDs): 0.21 mm, 0.55 mm, 0.21 mm; largest deviations: 0.6 mm, 1.0 mm, 0.5 mm] respectively. The larger deviation for the y component can be partly attributed to the CT slice thickness of 1 mm in that direction. Dose to the patient depends on the machine settings and patient geometry. To monitor dose consistency, air kerma (output) and half‐value layer (beam quality) are measured for a typical clinical setting. Air kerma was 6.3 cGy (120 kVp, 40 mA, 40 ms per frame, 360‐degree scan, S20 field of view); half value layer was 7.1 mm aluminum (120 kV, 40 mA).We suggest performing items 1, 2, and 3 monthly, and 4 and 5 annually. In addition, we devised a daily QA procedure to verify agreement of the megavoltage and kilovoltage isocenters using a simple phantom containing three small steel balls. The frequency of all checks will be reevaluated based on data collected during about 1 year.PACS number: 87.53.Xd