This article describes a pilot project of automatic morphosyntactic analysis system development and some results of our work. The approach, developed and adopted in this research, is caused by peculiarities of Arabic morphology and syntax, and implies parsing of morphosyntactic structures (both morphology and syntax) instead of traditional tokenization and division of language into morphology and syntax, which seems considerably artificial for many widespread phenomena of Arabic and leads to problems with parsing. Some existing Arabic corpora are discussed, some of them being even treebanks, but none of them having a common reliable underlying uniform formal model (a formal grammar of any kind) in public domain. The methodology of this project is described, and techniques used in the pilot study are discussed, including the software technologies adopted and developed. Current results are described with examples of morphosyntactic structures, immediate constituent classes (with information on dependencies), and code snippets of the grammar module.