mempalace-wanderer

You can build a local knowledge graph for agent memory by using mempalace_kg_add to seed facts with a layered triple pattern, then retrieving them through mempalace_kg_query. The workflow supports temporal fact invalidation to keep your local memory accurate.

Semantic search for local memory uses mempalace_search to recall relevant past content from local-first storage. The agent queries this memory before answering to retrieve historical context and prevent hallucinated facts during conversations.

Yes, you can create cross-domain connections in a temporal knowledge graph by using mempalace_create_tunnel to bridge wings and rooms. You verify these paths with mempalace_follow_tunnels for cross-domain pathfinding across distinct memory domains.

Yes, an MCP server setup is required for local-first semantic memory retrieval. You must configure the MemPalace MCP server and activate YAML-discovered tools like mempalace_status and mempalace_search before loading the agent skill.

A limitation of local memory palace persistence is the risk of hallucinated facts if the query-before-answer workflow is skipped. Users must also avoid common pitfalls like .hermes dot-quarks during knowledge graph management operations.