MemPalace Skill

MemPalace (v3.3.6) is a local-first AI memory system. It stores conversation history and project content verbatim — no summarization, no paraphrasing, no external API calls by default. Retrieval uses hybrid BM25 + semantic search, achieving 96.6% R@5 on LongMemEval with zero LLM involvement.

When to Use This Skill

• Setting up persistent memory for Claude Code or another LLM agent
• Mining project files (mempalace mine) or Claude Code transcripts (--mode convos)
• Retrieving forgotten context from past sessions (mempalace search, wake-up)
• Wiring auto-save hooks so sessions survive context compression
• Installing and configuring the MCP server (29 palace tools)
• Managing the knowledge graph (entities, predicates, temporal relationships)
• Backfilling existing ~/.claude/projects/ JSONL transcripts
• Repairing, deduplicating, or migrating a corrupted or upgraded palace
• Writing custom storage backends or switching embedding models
• Reproducing or extending memory benchmarks

Core Architecture