Neo4j Agent Memory MCP Server is a specialized MCP server that bridges a Neo4j graph database with AI agents to provide memory-centric tooling for storing, recalling, and connecting information in a knowledge graph. It exposes a set of atomic memory tools including search_memories, create_memory, create_connection, update_memory, update_connection, delete_memory, delete_connection, list_memory_labels, and get_guidance. The server enables persistent memory across conversations, semantic relationships (KNOWS, WORKS_AT, LIVES_IN, CREATED, etc.), natural language–style search across memory properties and relationships, and flexible labeling for memories (person, place, project, idea, etc.). Temporal tracking with automatic timestamps and date-based queries, plus graph exploration to traverse connections, supports context-aware depth-based searches. Enterprise-ready features allow connecting to multiple Neo4j databases, with an explicit NEO4J_DATABASE option. The design follows an LLM-Driven philosophy: no hidden logic, the LLM decides interpretation and relationships, and all actions are explicit and predictable, using simple atomic tools to enable complex workflows.