The DiffBind package in Bioconductor will calculate correlation values and plot clustered correlation heatmaps (as well as PCA plots, which are are useful to assess similarity amongst ChIP-seq samples).

This works by looking at all the peaks (called by MACS2 in your case), forming a consensus set (default includes peaks overlapping in at least two samples), and calculating correlation values. You can get a more accurate view by also supplying the aligned reads (bam files) and re-counting the number of reads overlapping each condensus peak site in each sample (regardless of whether it was called as a peak in that sample).