This paper presents CAmI, an agent oriented programming language for the development of ambient intelligence applications and for the further integration of the developed applications into complex, large-scale, AmI environments. The presented language is easy to learn and understand, modular and it offers many advantages that make it suited for the field of AmI, like the agent-oriented approach used, an intrinsic representation and use of the context, integration of various knowledge representation formalisms, independence from any computing platform, integration with function libraries written in other languages and, most important, features that facilitate the emergence of collective intelligence in the AmI environments, built in a bottom-up manner using CAmI.