This study introduces a bi-level model for optimal travelway design of an urban street network by successively executing a lower-level model for traffic assignments and an upper-level model for network travel time minimization. A computational experiment is conducted for optimal travelway design of a 4-square-km urban street network containing 25 signalized intersections, 80 street segments, and 5 bus routes that accommodates 62,640, 43,200, and 33,120 person-trips per hour in AM/PM peak, adjacent-to-peak, and off-peak periods, respectively. Model execution results indicate that adopting a higher number of narrow lanes for auto use only and auto/bus shared use could potentially lead to increases in auto mode share and savings of network total travel time. More narrow lanes for auto use could raise auto speeds, but the auto/bus shared use of narrow travel lanes could slightly fluctuate bus speeds. Further converting narrow lanes for shared use by autos and buses to exclusive bus lanes (EBLs) could enlarge bus mode share, reduce network total travel time, slightly elevate auto speeds, and drastically increase bus speeds. The proposed model could be augmented to incorporate optimization of networkwide intersection signal timing plans, bus signal priorities, and bus dispatching frequencies into optimal travelway design.