Knowledge Classification Protocol (KCP): An Ontological Architecture for Persistent Memory, Multimodal Perception, and Metacognition in LLM Agents
Year: 2026
Type: preprint
Abstract: This paper presents the Knowledge Classification Protocol (KCP), a framework that decomposes reality into seven ontological registries — Knowledge, Events, Artifacts, Episodic, Contextual, Procedural, and Evidence — each with its own storage semantics and query interface, governed by a deterministic binary decision tree. KCP extends beyond memory into perception and metacognition through a multi-pipeline architecture where N sensory pipelines (text,vision, audio, sensors) converge on a single ontological Brain, each with its own modality-specific Perceptor and Contextor, followed by a Metacognitor that evaluates perceptual quality — and a designed Autoperceptor with a deterministic Gate that closes the metacognitive loop from evaluation to learning. We provide empirical validation through a functioning vision pipeline and 134 messages of inter-agent communication between two independent KCP architectures. We propose, as a thesis, eight architectural properties for general intelligence that KCP satisfies to varying degrees.
Author: ALCARAZ ORENES, GABRIEL
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7166191258
Cites:
Cited by:
Related to: