We constrain the mass--richness scaling relation of redMaPPer galaxy clusters identified in the Dark Energy Survey Year 1 data using weak gravitational lensing. We split clusters into \(4\times3\) bins of richness \(\lambda\) and redshift \(z\) for \(\lambda\geq20\) and \(0.2 \leq z \leq 0.65\) and measure the mean masses of these bins using their stacked weak lensing signal. By modeling the scaling relation as \(\langle M_{\rm 200m}|\lambda,z\rangle = M_0 (\lambda/40)^F ((1+z)/1.35)^G\), we constrain the normalization of the scaling relation at the 5.0 per cent level as \(M_0 = [3.081 \pm 0.075 ({\rm stat}) \pm 0.133 ({\rm sys})] \cdot 10^{14}\ {\rm M}_\odot\) at \(\lambda=40\) and \(z=0.35\). The richness scaling index is constrained to be \(F=1.356 \pm 0.051\ ({\rm stat})\pm 0.008\ ({\rm sys})\) and the redshift scaling index \(G=-0.30\pm 0.30\ ({\rm stat})\pm 0.06\ ({\rm sys})\). These are the tightest measurements of the normalization and richness scaling index made to date. We use a semi-analytic covariance matrix to characterize the statistical errors in the recovered weak lensing profiles. Our analysis accounts for the following sources of systematic error: shear and photometric redshift errors, cluster miscentering, cluster member dilution of the source sample, systematic uncertainties in the modeling of the halo--mass correlation function, halo triaxiality, and projection effects. We discuss prospects for reducing this systematic error budget, which dominates the uncertainty on \(M_0\). Our result is in excellent agreement with, but has significantly smaller uncertainties than, previous measurements in the literature, and augurs well for the power of the DES cluster survey as a tool for precision cosmology and upcoming galaxy surveys such as LSST, Euclid and WFIRST.