Current AI agent architectures have made remarkable progress in what agents can do — connecting to tools, executing multi-step tasks, reasoning through complex problems, and managing context across long interactions. What has not been addressed is what agents cannot sense and do not hold: perception of the human's state, values that guide behavior beyond rule-following, adaptive routing between capabilities, self-awareness of their own performance and limits, and boundaries that define where the agent ends and the human begins. This paper introduces the Organic Agent Framework (OAF), a new paradigm for AI agent architecture modeled on biological organisms rather than software toolchains. OAF proposes that a complete agent requires not just a brain and hands, but a full body — an n-dimensional architecture comprising a moral codex foundation, modular sensing and processing organs, a body surface with protective security membrane, external wrappers for capability extension, and domain specialization as an environmental envelope. The paper presents the gap in current approaches, introduces the organism model and its core architectural primitives — including receiver-shaped inter-organ communication, variable-resolution perception, input contracts, controlled deception for data protection, and an inter-tier ingestion pattern — and demonstrates the architecture through the first fully defined organ. As agents grow more capable, what they lack becomes not less dangerous but more. OAF addresses the missing body.
Libor Klimeš (Tue,) studied this question.