Long-term memory
Also known as: persistent memory
Long-term memory is what an agent retains across sessions: persistent, retrievable memories of facts, decisions, and outcomes.
- Survives restarts and new conversations.
- Usually embeddings + a database (vector or key–value).
- Lets an agent build on past work instead of starting cold.
The value: an agent with long-term memory is a *continuity* — it remembers your conventions, your decisions, your mistakes. That's what makes it feel like a collaborator rather than a chatbot.