Memory-LanceDB FAQs

Question 1

What is Memory-LanceDB?

Accepted Answer

Memory-LanceDB is a standalone Model Context Protocol (MCP) server designed to give Claude Code a persistent, intelligent long-term memory system. It enables your AI assistant to remember past preferences, decisions, and project context across sessions.

Question 2

How does Memory-LanceDB prevent AI from forgetting?

Accepted Answer

It solves the AI 'forgetting' problem through automatic capture of important information after Claude Code's responses (Auto-Capture) and injecting relevant context at the start of new sessions (Auto-Recall). This ensures your AI assistant truly learns and retains knowledge over time, reducing repetition.

Question 3

What are the key memory capabilities of Memory-LanceDB?

Accepted Answer

Core capabilities include Hybrid Retrieval (vector similarity + BM25 full-text search with RRF fusion), LLM-powered Smart Extraction (6-category classification), and Intelligent Forgetting (using a Weibull decay model to fade irrelevant memories while preserving important ones).

Question 4

How does Memory-LanceDB integrate with Claude Code?

Accepted Answer

Memory-LanceDB runs as an independent MCP server, communicating with Claude Code via a standard JSON-RPC interface. Users can configure optional Claude Code hooks (e.g., SessionStart, Stop) for seamless, automatic memory capture and recall during their coding workflow.

Question 5

What types of information does it remember, and how is it organized?

Accepted Answer

It uses LLM-powered smart extraction to classify and remember various information types like preferences, decisions, facts, entities, events, and patterns. Memories are organized with Multi-Scope Isolation (agent-private, global, project-scoped) for flexible and relevant recall.

Memory-LanceDB

Memory-LanceDB

主な機能

ユースケース

主な機能

ユースケース