In Kyoto in 1997, the US government agreed that between 2008 and 2012 it would limit average annual emissions of greenhouse gases (GHGs) to seven percent below 1990 levels. As participants in the climate policy debate consider various means by which limits on US GHG emissions might be undertaken in the wake of the Kyoto agreement, there is considerable interest but also some confusion about how a GHG trading program could be organized and operated in practice. In this paper we address several aspects of policy design for a US system, such as who and what is covered by regulation, the organization of the trading system, how carbon permits are allocated, and how a system could be initiated and changed over time. The paper synthesizes existing analyses and adds new insights concerning uncertainty, intertemporal consistency, market institutions, and interactions with the tax system. Our fundamental conclusion is that a domestic "cap-and-trade" system with homogeneous permits applied to control flows of fossil fuels "upstream" in the energy system (along with selective inclusion of other gases and CO2 "sinks"), with permits auctioned periodically by the government, has the most appeal of different trading systems on efficiency and distributional grounds, though it may suffer politically because of its close resemblance to a carbon tax. We identify auction mechanisms that appear to be feasible and efficient for carbon permit allocation. We further argue that while the private sector should bear the "external" risk of changes in total permit availability as a consequence of modifications in international agreements, and that an auctioned upstream program provides more protection against the "internal" risk of efficiency-reducing opportunism by government regulators than other trading mechanisms.