While today's game engines and multi-agent platforms cross-fertilize each other to some extent, the technologies used in these areas are not readily compatible due to some differences in their primary concerns. Where game engines prioritize efficiency and central control, multi-agent platforms focus on agent autonomy and sophisticated communication capabilities. This volume gives an overview of the current state of the art for people wishing to combine agent technology with (serious) games. This state-of-the-art survey contains a collection of papers presented at AGS 2010;  the Second International Workshop on Agents for Games and Simulations, held on May 10, 2010, in Toronto, as well as extended versions of papers from other workshops and from the AAMAS conference. The 14 papers are organized in three topical sections focusing on architectures combining agents and game engines, on the training aspects of the games, on social and organizational aspects of games and agents, respectively.