memory-graph

Persistent memory stores durable facts, entities, and relationships in a knowledge graph so agents retain context between sessions without re-prompting. This Skill creates and manages that graph using the MCP Memory server, automatically capturing new discoveries like service owners, API endpoints, and component dependencies.

A knowledge graph organizes entities (people, services, endpoints) and their relationships (ownership, dependencies) into searchable nodes. This Skill provides tools to create entities, add observations with sources, establish relations, and prune outdated information—ensuring deduplication and atomic, idempotent operations.

Yes. This Skill proactively identifies emerging durable facts—requirements, decisions, owners, IDs, endpoints—and suggests storing them in memory. It records atomic observations with optional sources and routes updates via defined playbooks, so critical context is never lost.

The knowledge graph indexes relationships by entity type and attributes, allowing fast search and retrieval. You interact via MCP Memory tools (search_nodes, read_graph, open_nodes) to query connections and navigate the graph without manual reconstruction each session.

Prune when facts become stale, services sunset, or ownership changes. This Skill supports safe, idempotent delete operations with clear error handling, letting you maintain accuracy and reduce noise in stored memory over time.

Yes. The Skill integrates via fully-qualified MCP Memory tool names (mcp__memory__create_entities, add_observations, create_relations, delete_entities, read_graph, search_nodes, open_nodes), requiring only the @modelcontextprotocol/server-memory dependency.