The use of wireless multi-hop networks for disaster recovery systems and local area services is becoming a reality. However, cost-effective deployments of such networks still remain unfulfilled. In this work, we identify two important and previously unexplored deployment planning problems for such (wireless) multi-hop networks. The goal of these problems is to establish cost-optimized multi-hop paths between end-user devices by setting up infrastructure nodes at appropriate locations. A subtle and important practical constraint, which we call as the device-routing constraint, in choosing such paths requires that no end-user device becomes an intermediate node on any path. With cost on network edges as well as network nodes, this problem of choosing appropriate locations is in contrast to the traditional mesh network planning, backbone routing, or facility location problems. We show that the problem is NP-hard, and that the network planning algorithms in prior work are inapplicable for networking planning with the device-routing constraint. We then present Host-Bypass, the first approximation algorithm for these problem with provable performance bound on cost. Host-Bypass has polynomial running time and our simulation study shows that Host-Bypass performs 30%better than a heuristic from a well-known algorithm in prior work.