Collaborative procurement has been considered as a vital link in many world-class supply chain strategies. Most of the previous studies or real-cases related to the e-marketplace or auction based collaborative procurement just covered the purchasing of distinct items. However, each buyer may have the purchase needs for multiple heterogeneous items at the same time. Likewise, large scale distributors or manufactures or who would like to sell items may have the combinatorial supply needs of selling two or more items simultaneously according to their situation. Therefore, in this research, we address a multi-agent based combinatorial auction model for collaborative procurement as a new business model that combines the strength of the combinational auction and the benefits of collaborative procurement. A representative combinatorial auction based collaborative procurement (CACP) process is presented, then, the system architecture and optimized bid selection modeling for CACP are described. Finally we show the viability of CACP through proof-of-concept prototype implementation and experiment results.