The South African grassland biome is one of the most threatened biomes in South Africa. Approximately 45% of the grassland biome area is transformed, degraded or severely invaded by alien plants and the remaining natural areas are highly fragmented. In this fragmented landscape, the connectivity between habitat patches is very important to maintain viable populations. In this study we aimed to quantify connectivity of the grassland biome in Mpumalanga using graph theory in order to identify conservation priorities and to direct conservation efforts. Graph theory-based connectivity indices have the ability to combine spatially explicit habitat data with species specific dispersal data and can quantify structural and functional connectivity over large landscapes. We used these indices to quantify the overall connectivity of the study area, to determine the influence of abandoned croplands on overall connectivity, and to identify the habitat patches and vegetation types most in need of maintaining overall connectivity. Natural areas were identified using 2008 land cover data for Mpumalanga. Connectivity within the grassland biome of Mpumalanga was analysed for grassland species with dispersal distances ranging from 50 to 1000 m. The grassland habitat patches were mostly well connected, with 99.6% of the total habitat area connected in a single component at a threshold distance of 1000 m. The inclusion of abandoned croplands resulted in a 33% increase in connectivity at a threshold distance of 500 m. The habitat patches most important for maintaining overall connectivity were the large patches of continuous habitat in the upper and lower centres of the study area and the most important vegetation types were the Wakkerstroom Montane Grassland and the Eastern Temperate Freshwater Wetlands. These results can be used to inform management decisions and reserve design to improve and maintain connectivity in this biome.