Maintaining ecosystem processes within patches of remnant vegetation is critical to minimising biodiversity loss in agricultural landscapes. Foundation species—habitat-forming organisms that interact with many other species—are therefore a priority for conservation and restoration in farming areas. Triodia spp. grasses are foundation species of arid and semi-arid Australia that largely depend on fire for recruitment, but fire is suppressed or excluded in many agricultural areas. We tested the effectiveness of controlled burns and competitor removal (exotic and native grasses), both in isolation and combined, on recruitment rates of Triodia scariosa in remnant vegetation using a before-after, control-impact study across 126 plots. A subset of plots were located in an herbivore exclusion area inside a reserve. There was no recruitment of T. scariosa 1 year after burning, regardless of treatment or control, and the burns killed all existing plants. However, T. scariosa germinated by 2 years post-burn, with the greatest recruitment in sites where both burning and grass removal were applied. Two years after burning, T. scariosa abundance (adults and recruits) remained low outside reserves, but returned to original levels in reserves and in areas where large herbivores were excluded. Synthesis and applications. Despite failing to increase overall abundance, we show that restoration of a foundation plant species in degraded, agricultural landscapes can be achieved through a combination of fire reintroduction and competitor suppression. Germination trials from soil samples suggest a depleted seed bank limited recruitment rates, and therefore emphasise the importance of carefully timing restoration actions to overcome recruitment bottlenecks.