Test paper generation for examination in various level of education is one of the most preliminary requirements. Preparing and processing a test paper has vital place at every stage of E-education system. Most subsisting systems automate test paper generation by randomly selecting question items from knowledgebase/database. Usually these systems have no concern about difficulty level of question items. Some intelligent systems have been developed. But mostly it considers difficulty level at the time of assessment. Some other intelligent system gives approximate solutions for producing a test paper while considering difficulty level and type etc.  Unlike the existing test paper generators, our utility based test paper agent chooses question items in such a way that the difficulty level of each question item takes part in computing exact difficulty level of test paper. Proposed Utility based test paper agent (UBTPAgent) picks the question items with its utility value. Hence provides exact difficulty level for whole paper as required by user. So test papers on the same level are different on basis of culled difficulty level. In design phase examiner creates a knowledgebase of questions for UBTPAgent by assigning some utility values with every question on which our selection algorithm operates. Whenever a test paper is required examiner provides difficulty level for test paper. And on the behalf of examiner UBTPAgent selects question with utility value in such a way the total of utility value is equal to required difficulty level. Later, a test paper is produced by test paper generator according to difficulty level specified by examiner. As percentage for any test paper is 1 to 100 percent for any test paper, therefore, 100 difficulty levels are available for any test paper. Here, selected difficulty level is for the entire paper. Finally, UBTPAgent model is proposed, implemented by providing algorithm, executed using case study, and tested to ensure the feasibility of this approach.