The James Webb Space Telescope (JWST) is expected to revolutionize our understanding of the high-redshift Universe, and may be able to test the prediction that the first, chemically pristine (population III) stars formed with very high characteristic masses. Since isolated population III stars are likely to be beyond the reach of JWST, small population III galaxies may offer the best prospects of directly probing the properties of metal-free stars. Here, we present Yggdrasil, a new spectral synthesis code geared towards the first galaxies. Using this model, we explore the JWST imaging detection limits for population III galaxies and investigate to what extent such objects may be identified based on their JWST colours. We predict that JWST should be able to detect population III galaxies with stellar population masses as low as ~10^5 Msolar at z=10 in ultra deep exposures. Over limited redshift intervals, it may also be possible to use colour criteria to select population III galaxy candidates for follow-up spectroscopy. The colours of young population III galaxies dominated by direct star light can be used to probe the stellar initial mass function (IMF), but this requires almost complete leakage of ionizing photons into the intergalactic medium. The colours of objects dominated by nebular emission show no corresponding IMF sensitivity. We also note that a clean selection of population III galaxies at z~7-8 can be achieved by adding two JWST/MIRI filters to the JWST/NIRCam filter sets usually discussed in the context of JWST ultra deep fields.