G OAL: A Multi-Agent Programming Language Applied to an Exploration Game Koen V. Hindriks and J¨ urgen Dix Abstract GOAL is a multi-agent programming language based on the BDI paradigm. It is a logic-based language that supports modular agent design based on established software engineering principles and interaction with environments using an environ- ment interface standard (EIS). GOAL recently won the Multi-Agent Programming Contest (MAPC), where two teams consisting of ten agents play against each other in order to explore and defend occupied territory on a distant planet. The MAPC game is a complex and dynamic environment that supports EIS and thus facilitates easy connection of a multi-agent system to an environment that is remotely run. We describe the design of the multi-agent solution that won the competition, the EIS interface that was used, and the MAPC scenario. 1 Introduction The aim of this chapter is not to describe yet-another agent programming language and claim that it is the best on the market. Developing good software for non-trivial applications using the agent paradigm is a highly complex task depending not only on the chosen programming language. We strongly believe that documenting and discussing projects that use existing agent platforms for software development is useful for a number of reasons. Only by actually using such platforms can we learn about the effectiveness and usability of them as well as about the issues we are facing during such projects. Based upon findings related to the development process itself, comments by software develop- Koen V. Hindriks Delft University of Technology, Mekelweg 4, Delft, The Netherlands, e-mail: k.v.hindriks@ tudelft.nl J¨ urgen Dix Clausthal University of Technology, Julius-Albert-Str. 4, Clausthal-Zellerfeld, Germany, e-mail: dix@tu-clausthal.de 1