Over the last decades, cosmogenic exposure dating has permitted major advances in many fields of Earth surface sciences and particularly in paleoglaciology. Yet, exposure age calculation remains a complicated and dense procedure. It requires numerous choices of parameterization and the use of an accurate production rate. This study describes the CREp program and the ICE-D production rate online database. This system is designed so that the CREp calculator will automatically reflect the current state of this global calibration database production rate, ICE-D. ICE-D will be regularly updated in order to incorporate new calibration data and reflect the current state of the available literature. CREp is a Octave/Matlab© online code that computes Cosmic Ray Exposure (CRE) ages for 3He and 10Be, available at crep.crpg.cnrs-nancy.fr. A stand-alone version of the CREp code is also released with the present article. Note however that only the online version is connected to the online database ICE-D. The CREp program offers the possibility to calculate ages with two scaling models: i.e. the empirical Lal-Stone time-dependent model (Balco et al., 2008; Lal, 1991; Stone, 2000) with the muon parameters of Braucher et al. (2011), and the Lifton-Sato-Dunai (LSD) theoretical model (Lifton et al., 2014). The default atmosphere model is the ERA-40 database (Uppala et al., 2005), but one may also use the standard atmosphere for comparison (N.O.A.A, 1976) to apply the atmospheric correction. To perform the time-dependent correction, users may import their own geomagnetic database for paleomagnetic corrections or opt for one of the three proposed datasets (Lifton, 2016; Lifton et al., 2014; Muscheler et al., 2005). For the important choice of the production rate, CREp is linked to a database of production rate calibration data that is part of the ICE-D (Informal Cosmogenic-nuclide Exposure-age Database) project (http://calibration.ice-d.org). This database includes published empirical calibration rate studies that are publicly available at present, comprising those of the CRONUS-Earth and CRONUS-EU projects, as well as studies from other projects. In the present study, the efficacy of the different scaling models has also been evaluated looking at the statistical dispersion of the computed Sea Level High Latitude (SLHL) production rates. Lal/Stone and LSD models have comparable efficacies, and the impact of the tested atmospheric model and the geomagnetic database is also limited. Users however have several possibilities to select the production rate: 1) using a worldwide mean value, 2) a regionally averaged value (not available in regions with no data), 3) a local unique value, which can be chosen among the existing dataset or imported by the user, or 4) any combination of single or multiple calibration data. If a global mean is chosen, the 1σ uncertainty arising from the production rate is about 5% for 10Be and 10% for 3He. If a regional production rate is picked, these uncertainties are potentially lower. CREp is able to calculate a large number of ages in a reasonable time (typically