We study the potential of the galaxy cluster sample expected from the Chinese Space Station Telescope (CSST) survey to constrain dark energy properties. By modeling the distribution of observed cluster mass for a given true mass to be log-normal and adopting a selection threshold in the observed mass M200m ≥ 0.836 × 1014 h−1 M⊙, we find about 4.1 × 105 clusters in the redshift range 0 ≤ z ≤ 1.5 can be detected by the CSST. We construct the Fisher matrix for the cluster number counts from CSST, and forecast constraints on dark energy parameters for models with constant (w0CDM) and time dependent (w0waCDM) equation of state. In the self-calibration scheme, the dark energy equation of state parameter w0 of the w0CDM model can be constrained to Δw0 = 0.036. If wa is added as a free parameter, we obtain Δw0 = 0.077 and Δwa = 0.39 for the w0waCDM model, with a Figure of Merit for (w0, wa) of 68.99. Should we have perfect knowledge of the observable-mass scaling relation ("known SR" scheme), we would obtain Δw0 = 0.012 for the w0CDM model, and Δw0 = 0.062 and Δwa = 0.24 for the w0waCDM model. The dark energy Figure of Merit of (w0, wa) increases to 343.25. This indicates again the importance of calibrating the observable-mass scaling relation for optically selected galaxy clusters. By extending the maximum redshift of the clusters from \(z_{\rm max}\sim 1.5\) to \(z_{\rm max}\sim 2\), the dark energy Figure of Merit for (w0, wa) increases to 89.72 (self-calibration scheme) and 610.97 ("known SR" scheme), improved by a factor of ∼1.30 and ∼1.78, respectively. We find that the impact of clusters' redshift uncertainty on the dark energy constraints is negligible as long as the redshift error of clusters is smaller than 0.01, achievable by CSST. We also find that the bias in logarithm mass must be calibrated to be 0.30 or better to avoid significant dark energy parameter bias.