Synaptic plasticity is widely found in many areas of the central nervous system. In particular, it is believed that synaptic depression can act as a mechanism to allow simple networks to generate a range of different firing patterns. The simplicity of the locomotor circuit in hatchling Xenopus tadpoles provides an excellent place to understand such basic neuronal mechanisms. Depending on the nature of the external stimulus, tadpoles can generate two types of behaviours: swimming when touched and slower, stronger struggling movements when held. Struggling is associated with rhythmic bends of the body and is accompanied by anti-phase bursts in neurons on each side of the spinal cord. Bursting in struggling is thought to be governed by a short-term synaptic depression of inhibition. To better understand burst generation in struggling, we study a minimal network of two neurons coupled through depressing inhibitory synapses. Depending on the strength of the synaptic conductance between the two neurons, such a network can produce symmetric n−n anti-phase bursts, where neurons fire n spikes in alternation, with the period of such solutions increasing with the strength of the synaptic conductance. Using a fast/slow analysis, we reduce the multidimensional network equations to a scalar Poincaé burst map. This map tracks the state of synaptic depression from one burst to the next, and captures the complex bursting dynamics of the network. Fixed points of this map are associated with stable burst solutions of the full network model, and are created through fold bifurcations of maps. We prove that the map has an infinite number of stable fixed points for a finite coupling strength interval, suggesting that the full two-cell network also can produce n−n bursts for arbitrarily large n. Our findings further support the hypothesis that synaptic depression can enrich the variety of activity patterns a neuronal network generates.