Intelligent agents in AI are entities that perceive their environment, take autonomous actions, and use tools to accomplish user-defined tasks in various environments.

Their capabilities and effectiveness depend on available tools and their planning mechanisms, with complex, multi-step operations introducing higher chances of error and increased resource use. Agents can operate software, web, or physical environments, and may use additional tools for knowledge augmentation, capability extension, and environmental interaction.

The current field lacks unified theoretical frameworks, and practical implementation remains experimental and rapidly evolving.