Joint cost management is a decisive factor for sustainable collaboration in supplier-buyer dyads. For both parties the establishment of an accurate cost estimation (CE) framework supports managing suppliers' costs as well as manufacturer's quotation costing. Grounded on resource dependence theory and following a design science research approach, this study introduces a multi-perspective CE system inspired by statistical learning, deep learning, decision making, and multi-agent theory. We evaluate our system by a single case and computer simulation study, using empirical data coming from observations and archives at a large Bavarian original equipment manufacturer (OEM). The results indicate that our CE approach allows to select the most significant cost-drivers and predict total costs of parts and assemblies with high accuracy. This supports the supplier in efficiently managing its costs. In making the CE blackbox model transparent using a combination of model agnostic post-hoc explainable artificial intelligence approaches we foster user acceptance for both suppliers and OEMs. All CE artifacts are ensembled in a multi-agent system to automatically manage costs with suppliers and furthermore, as a model extension, can lead to a collaborative price agreement. Our system supports supply chain managers on both sides in entering into a sustainable long partnership.