A world wide investigation for the various aspects of machine translation between natural languages is underway. In this work, the Urdu grammar frame work based on Lexical Functional Grammar (LFG) theory is presented that preserves the grammatical relations such as gender, number, tense, etc. across various levels of syntactic representations. The proposed framework takes care of verb agreement transitive and in-transitive verbs, complex predicates, case markers and scrambling. The transfer between languages is done at f-structure level, where syntactic difference between languages is at its minimum and thus that enables successful grammatical machine translation.