Long-Term Memory FAQs

Question 1

Can I see my AI's memory graph in real-time?

Accepted Answer

Yes, Long-Term Memory includes a real-time WebSocket-based visualization of the memory graph. This live view shows memories being saved, searched, or deleted, offering transparency into how your AI processes and connects information.

Question 2

Is Long-Term Memory easy to integrate with existing AI frameworks?

Accepted Answer

Absolutely. It's designed as a drop-in MCP server, offering quick setup for MCP clients like Claude Desktop, Claude Code, and OpenClaw. It automatically injects memory usage instructions into your client's configuration files for seamless integration.

Question 3

How does Long-Term Memory address token waste and manual retrieval?

Accepted Answer

It solves manual retrieval by automatically injecting relevant memories with its `auto_search` feature. For token waste, it employs multi-resolution text composition to select only the most relevant memory summaries or triples within a token budget, optimizing context usage.

Question 4

What is Long-Term Memory for AI assistants?

Accepted Answer

Long-Term Memory provides AI assistants, like Claude, with persistent and self-organizing memory capabilities. It enables them to remember information across conversations, utilizing advanced techniques to store, retrieve, and manage context intelligently.

Question 5

What advanced retrieval methods does Long-Term Memory use?

Accepted Answer

It features a hybrid retrieval system combining semantic search (ChromaDB with multilingual embeddings) and knowledge graph traversal. An RL-powered re-ranker fuses and prioritizes results, while a contextual bandit policy intelligently decides when to save, skip, or retrieve memories.

Long-Term Memory

Long-Term Memory

Key Features

Use Cases

Key Features

Use Cases