Kinetics of nutrient degradation in the rumen is an important component of feed evaluation systems for ruminants. The in situ technique is commonly used to obtain such dynamic parameters, but it requires cannulated animals and incubations last several days limiting its application in practice. On the other hand, feed industry relies strongly on NIRS to predict chemical composition of feeds and it has been used to predict nutrient degradability parameters. However, most of these studies were feedstuff specific, predicting degradability parameters of a particular feedstuff or category of feedstuffs, mainly forages or compound feeds and not grains and byproducts. Our objective was to evaluate the potential of NIRS to predict degradability parameters and effective degradation utilizing a wide range of feedstuffs commonly used in ruminant nutrition. A database of 809 feedstuffs was created. Feedstuffs were grouped as forages (FF; n = 256), non-forages (NF; n = 539) and of animal origin (n = 14). In situ degradability data for dry matter (DM; n = 665), CP (n = 682) and NDF (n = 100) were collected. Degradability was described in terms of washable fraction (a), slowly degradable fraction ( b) and its rate of degradation (c). All samples were scanned from 1100 to 2500 nm using an NIRSystems 5000 scanning in reflectance mode. Calibrations were developed for all samples (ALL), FF and NF. Equations were validated with an external validation set of 20% of total samples. NIRS equations to predict the effective degradability and fractions a and b of DM, CP and NDF could be evaluated from being adequate for screening (r 2 > 0.77; ratio of performance to deviation (RPD) = 2.0 to 2.9) to suitable for quantitative purposes (r 2 > 0.84; RPD = 3.1 to 4.7), and some predictions were improved by group separation reducing the standard error of prediction. Similarly, the rate of degradation of CP (CP c ) and DM (DM c ) was predicted for screening purposes (RPD ⩾2 and 2.5 for CP c and DM c , respectively). However, the rate of degradation of NDF was not predicted accurately (NDF c : r 2 < 0.75; RDP < 2).