Hi Sobia,

the answer depends on the software you are using. For Stata, some commands have the option of multi-dimensional clustering "built in". If you need to do it by hand, you can go to the website of Nathan Nunn (Harvard Econ) and download the Do-files for the article he wrote together with Leonard Wantchekon on the long-term effect of slavery on trust. Somewhere in there, they compute double clustered standard errors by hand (in their 2sls estimations). 

Hope this helps,

David