Connected and automated vehicles (CAVs) have real-time knowledge of the immediate driving environment, actions to be taken in the near future and information from the cloud. This knowledge, referred to as preview information, enables CAVs to drive safely, but can also be used to minimize fuel consumption. Such fuel-efficient transportation has the potential to reduce aggregate fuel consumption by billions of gallons of gas every year in the U.S. alone. In this paper, we propose a planning framework for use in CAVs with the goal of generating fuel-efficient vehicle trajectories. By utilizing on-board sensor data and vehicle-to-infrastructure (V2I) communications, we leverage the computational power of CAVs to generate eco-friendly vehicle trajectories. The planner uses an eco-driver model and a predictive cost-based search to determine the optimal speed profile for use by a CAV. To evaluate the performance of the planner, we introduce a co-simulation environment consisting of a CAV simulator, Matlab/Simulink and a CAV software platform called the InfoRich Eco-Autonomous Driving (iREAD) system. The planner is evaluated in various urban traffic scenarios based on real-world road network models provided by the National Renewable Energy Laboratory (NREL). Simulations show an average savings of 14.5% in fuel consumption with a corresponding increase of 2% in travel time using our method.