Context Graphs

Context graphs are graph-backed memory and retrieval structures that connect entities, events, policies, previous decisions, tool traces, and reasoning records so agents can make and explain decisions rather than merely retrieve documents ^[src-088].

Key points

• Neo4j's late-May AI Engineer talks frame context graphs as a decision layer above ordinary RAG: they supply the facts, precedents, policies, and relationships needed to answer "what should happen and why?" ^[src-088].
• A context graph can combine short-term conversation memory, long-term entity memory, and reasoning memory into one queryable structure ^[src-088].
• Decision traces matter because many production questions depend on precedents, causal chains, authority, reversibility, risk, and the cost of being wrong, not only semantic similarity to a document ^[src-088].
• Graph-native retrieval can mix Cypher queries, vector search, graph embeddings, and graph algorithms to find structurally similar cases as well as semantically similar text ^[src-088].
• The practical promise is auditability: humans can inspect which entities, policies, previous decisions, and risk factors the agent used before accepting, rejecting, or escalating a decision ^[src-088].

Related entities

• Neo4j

Related concepts

• Decision Traceability For Agents
• Context Engineering
• Enterprise Knowledge Graph
• Agentic Context Management
• LLM Observability
• Agent Forensics

Source references

• ^[src-088] AI Engineer late-May 2026 channel update (48 transcripts, 2026-05-15 to 2026-05-31)