ABSTRACT The Pristine survey uses narrow-band photometry to derive precise metallicities down to the extremely metal-poor regime ($ \rm Fe/H \lt -3$), and currently consists of over 4 million FGK-type stars over a sky area of $\sim 2500\, \mathrm{deg}^2$. We focus our analysis on a subsample of ∼80 000 main-sequence turn-off stars with heliocentric distances between 6 and 20 kpc, which we take to be a representative sample of the inner halo. The resulting metallicity distribution function (MDF) has a peak at $ \rm Fe/H =-1.6$, and a slope of Δ(LogN)/$\Delta \rm Fe/H = 1.0 \pm 0.1$ in the metallicity range of $-3.4\; \lt\; \rm Fe/H\; \lt -2.5$. This agrees well with a simple closed-box chemical enrichment model in this range, but is shallower than previous spectroscopic MDFs presented in the literature, suggesting that there may be a larger proportion of metal-poor stars in the inner halo than previously reported. We identify the Monoceros/TriAnd/ACS/EBS/A13 structure in metallicity space in a low-latitude field in the anticentre direction, and also discuss the possibility that the inner halo is dominated by a single, large merger event, but cannot strongly support or refute this idea with the current data. Finally, based on the MDF of field stars, we estimate the number of expected metal-poor globular clusters in the Milky Way halo to be 5.4 for $ \rm Fe/H\; \lt\; -2.5$ and 1.5 for $ \rm Fe/H\; \lt\; -3$, suggesting that the lack of low-metallicity globular clusters in the Milky Way is not due simply to statistical undersampling.