Context Graphs
Knowledge graphs encode structured facts about a domain — entities, attributes and relationships — and answer “what is true?” remarkably well. As systems become more autonomous, driven by large language models, agents and continuous streams of organisational activity, a different question dominates: “what is true right now, for this user, in this situation, and why?” This paper introduces and formalises the context graph, extending the knowledge graph with four dimensions: time, provenance, situational scope and decision traces. It presents the conceptual model, a reference architecture for building context graphs from live operational data, a systematic comparison with classical knowledge graphs, practical applications across RAG, proactive enterprise agents, compliance and personalisation, a pragmatic construction methodology, and open challenges around decay, privacy and scale.