I'm not quite sure what you're asking. There are two possibilities:
If you are using individual subject's original scans, and therefore a mask in subject space rather than MNI space (that maybe you've back projected using fnirt or flirt) then you can use fslstats with the -k and -M flags. You could still run fslmeants, you'll only get one subject's values that way, but at least you'll have your values from the cluster. You could write a little for loop to speed up the extraction.
If you have run a group analysis (eg: TBSS, VBM) and you want to extract values from a cluster that shows a significant results then you can run fslmeants on the (eg) all_FA.nii.gz or all_FA_skeletonised.nii.gz image. These are 4D files that have images for all participants and importantly they're all in the same space as the result file that you'll pass to the cluster command.
I hope this helps.