Abstract Mutualism is often conditional and the associations vary from antagonism to mutualism along environmental conditions and genotypes of interacting species. I studied antagonism-mutualism continuum hypothesis of symbiosis experimentally using two different Epichloë/Neotyphodium endophytes and their host grasses, agricultural meadow fescues and natural fine fescue, as study systems. These systemic fungal endophytes live asymptomatically within aerial tissues of grasses, and are vertically transmitted to the next grass generation via seeds. Thus, asexual endophyte strains are dependent on the growth, survival and reproduction of their hosts. Epichloë/Neotyphodium endophytes are considered plant mutualists, because they improve the resistance of the host against various stresses, e.g. herbivores. In addition to experimental approach, I examined prevalence and genetic structure of Epichloë festucae in natural grass populations. Finally, current knowledge concerning grass endophytes was reviewed and the effects of variable environment and genetic background on the ecology and the evolution of grass-endophyte symbiosis were discussed. The endophyte improved the performance of the agronomic meadow fescues, but the beneficial effects were dependent on the grass cultivar and the growth environment. The endophyte-infected (E+) meadow fescues were more susceptible to the pathogenic snow molds and they suffered increased winter damage compared to the endophyte-free (E-) plants. Many natural Festuca rubra and F. ovina populations were either endophyte-free or had low infection frequency. The highest infection frequencies were found in subarctic areas where the infection incidence differed between habitats. Twenty out of the 25 E. festucae genotypes detected were carrying multiple alleles in microsatellite loci indicating multiple infections or vegetative hybridization of the fungus. A dominant genotype (63.5% of all isolates) occurred in all populations suggesting that this fungus is mainly asexual. E+ F. ovina seedlings performed worse than endophyte-free E- seedlings. In F. rubra, the river bank originated E+ seedlings allocated fewer, but longer and heavier tillers than the other seedlings indicating possible improved performance of the endophyte infected grasses in harsh river bank conditions. In short, I detected both positive and negative effects of endophyte infection on grasses varying along species, environment and genotypic background of study subjects. The results support the antagonism-mutualism continuum hypothesis.