neo4j-agent-memory-skill

Persistent agent memory with Neo4j is implemented using the neo4j-agent-memory package to store conversation history, extracted entities, and reasoning traces as graph data for coherent retrieval. The Skill provides the authoritative reference for mapping MemoryClient and MemorySettings usage.

The POLE+O model for long-term agent memory structures extracted knowledge into Person, Object, Location, Event, and Organization entities derived from AI assistant interactions. It models structured knowledge as graph nodes within Neo4j for persistent retrieval.

Reasoning traces are stored as graph data in Neo4j by persisting thought, action, and provenance as first-class graph nodes. This allows AI agents to retrieve and learn from prior reasoning patterns during future interactions.

The Neo4j Agent Memory Service supports MCP tools through its hosted endpoint at memory.neo4jlabs.com. It provides REST and hosted MCP tools for managed memory operations and sharing, requiring authentication with a nams_ key prefix.

The Neo4j agent memory package supports short-term conversation history by storing context as graph data for continuity across sessions. This enables context retrieval for AI assistants interacting with the Neo4j graph database.

The hosted NAMS service requires correct mapping of MemoryClient and MemorySettings usage along with specific endpoint and authentication details, including the nams_ key prefix. Safe and accurate implementation depends on following these hosted service constraints strictly.