A barter system is an alternative commerce approach where customers meet at a marketplace in order to exchange their goods or services without currency. In order to cope with challenges of electronic barter systems (e.g. efficient management of trades and determination of best-matching goods and their suppliers), several software systems based on intelligent agents have been proposed. However only a very few of these proposals consider exact development and implementation of multi-agent e-barter systems. Besides, most of the approaches consider matching between exchanged goods in which only price and quantity information are used. Hence, in this paper, we discuss design and implementation of a multi-agent e-barter system which utilizes ontology-based comparison for bid matching. Formal representation and decision-making criteria for agent mediated barter process are given and behavioral model of collaborating agents within the system is described. In addition to the traditional e-barter members, a new type of software agent is introduced in order to infer about semantic closeness between offered and purchased items. Related approach may enhance capabilities of e-barter systems in the way of finding the most appropriate matches between supplies and demands, considering not only price and quantities for goods.