agent-wow is an open-source project exploring the capabilities of LLM-powered agents to play World of Warcraft autonomously.
Inspired by projects such as Mindcraft, agent-wow aims to contribute to the growing body of research into how effectively AI agents can operate within complex simulated worlds, make decisions over long time horizons, and work toward defined goals with minimal human intervention.
This blog documents a series of experiments with agent-wow, gradually challenging AI agents with increasingly complex tasks. Our ultimate goal is to fill an entire server with autonomous AI agents capable of progressing from level 1 all the way to defeating the Lich King on Heroic difficulty. Along the way, we hope to learn a great deal about the capabilities and limitations of current AI models in long-running, multiplayer environments.
Subscribe to the blog via email or RSS feed.
agent-wow is intended for research, educational, and testing purposes. All published experiments are conducted on private servers running AzerothCore. agent-wow communicates directly with the server using the WoW network protocol and does not alter any Blizzard client software or game assets.