Overview
MemPalace is the leading open-source, local-first AI memory system, engineered for verbatim storage and semantic retrieval of your conversation history. It boasts industry-leading benchmarks for recall accuracy—achieving 96.6% R@5 raw on LongMemEval without any API calls. This means your AI interactions are stored securely and privately on your machine, with a structured approach to organization that enhances searchability.
Key Features
- Verbatim Storage: MemPalace preserves your conversation history exactly as it happened, avoiding summarization or paraphrasing for maximum fidelity.
- Structured Indexing: Conversations are organized into a hierarchical "palace" structure with "wings" for people/projects, "rooms" for topics, and "drawers" for original content, allowing for highly scoped searches.
- Pluggable Backend: The retrieval layer is designed to be flexible, with ChromaDB as the default, enabling easy integration of alternative vector stores.
- Local-First & Private: All data remains on your machine, ensuring complete privacy and control. No data leaves your system unless you explicitly opt-in.
- Top-Tier Performance: Outperforms other memory systems with a raw retrieval recall of 96.6% R@5 on the LongMemEval benchmark, and can reach over 99% with LLM reranking.
- Knowledge Graph: Features a temporal entity-relationship graph for sophisticated data querying and management.
- MCP Server & Tools: Includes 29 MCP tools for comprehensive interaction with your AI memory, including cross-wing navigation and agent diaries.
- Agent Support: Enables each specialist agent to have its own distinct wing and diary within the palace, discoverable at runtime without bloating system prompts.
Typical Use Cases
- Enhanced AI Assistant Memory: Keep a detailed, searchable record of all your interactions with AI assistants like Claude, Gemini, or local models.
- Project Development & Collaboration: Store and retrieve crucial details from project discussions, code reviews, and planning sessions.
- Personal Knowledge Management: Build a private, semantic index of your learning materials, notes, and research.
- Agent Diary & State Management: Provide persistent memory and context for autonomous AI agents.
- Secure Data Archiving: Ensure verbatim, easily retrievable archives of sensitive or important conversations.
