Smart grazing is a relatively difficult field of digital agriculture. Restricted by the geographical conditions of pastures, poor network infrastructure and low economic output, relevant applications have been difficult to achieve. In this paper, we propose the SAG-GIoT system based on the background of yak grazing production in the Qinghai Tibet Plateau, and define three intelligent grazing management application scenarios: daily grazing supervision, UAV grazing and searching for yaks. To this end, we have designed the three-tier technical architecture of SAG-GIoT, and developed collar, base station and grazing management system. We designed the all-terrain network service scheme with BDS-BSS and SBSs, and verified the daily grazing supervision test in long-term. UAV grazing test was carried out in pasture, and a flexible communication networking was realized through the UAV-BS. With the guidance of UAV searching and APP positioning, taking HBSs in hand, we quickly and accurately find the lost yaks. SAG-GIoT system is characterized as low cost, flexible deployment and global service, and has broad application prospects.