AbstractFacilitation among plants mediated by grazers occurs when an unpalatable plant extends its protection against grazing to another plant. This type of indirect facilitation impacts species coexistence and ecosystem functioning in a large array of ecosystems worldwide. It has nonetheless generally been understudied so far in comparison with the role played by direct facilitation among plants. We aimed at providing original data on indirect facilitation at the community scale to determine the extent to which indirect facilitation mediated by grazers can shape plant communities. Such experimental data are expected to contribute to refining the conceptual framework on plant–plant–herbivore interactions in stressful environments. We set up a 2‐year grazing exclusion experiment in tropical alpine peatlands in Bolivia. Those ecosystems depend entirely on a few, structuring cushion‐forming plants (hereafter referred to as “nurse” species), in which associated plant communities develop. Fences have been set over two nurse species with different strategies to cope with grazing (direct vs. indirect defenses), which are expected to lead to different intensities of indirect facilitation for the associated communities. We collected functional traits which are known to vary according to grazing pressure (LDMC, leaf thickness, and maximum height), on both the nurse and their associated plant communities in grazed (and therefore indirect facilitation as well) and ungrazed conditions. We found that the effect of indirectly facilitated on the associated plant communities depended on the functional trait considered. Indirect facilitation decreased the effects of grazing on species relative abundance, mean LDMC, and the convergence of the maximum height distribution of the associated communities, but did not affect mean height or cover. The identity of the nurse species and grazing jointly affected the structure of the associated plant community through indirect facilitation. Our results together with the existing literature suggest that the “grazer–nurse–beneficiary” interaction module can be more complex than expected when evaluated in the field.