There are different Arabic corpora that are freely available online. However, most of these are experimental or in the developmental stage and none is of the magnitude of ones such as the BNC or COCA.
You can have a look at the Quranic Arabic Corpus (http://corpus.quran.com/) or a selection available here (http://www.comp.leeds.ac.uk/eric/latifa/research.htm)
So using one or the other all depends on whether your research focuses on general language or is more specific in scope. You can even make a comparison between concordance from different corpora.
Where the corpus is not searchable online, you will have to make use of a concordancer. Some of these are accessible through these links: AConcorde (http://www.andy-roberts.net/coding/aconcorde), AntConc (http://www.laurenceanthony.net/software/antconc/).
I also suggest giving the following tool a try: WebCorp (http://www.webcorp.org.uk/live/). It uses the net as a corpus and the benefits may outweigh the shortcomings of the corpus quality and structure.
There is a United Nations Arabic Corpus on BFSU CQPweb (http://www.bfsu-corpus.org/channels/corpus). You can also try Quranic Arabic Corpus (http://corpus.quran.com/) and arabiCorpus (http://arabicorpus.byu.edu/).
But it depends on your specific research question.