// ⌘ Agents & Tool Use
MemGPT (Hierarchical Agent Memory)
LLM manages its own memory like an OS – paging between the context window and external storage.
MemGPT 分层记忆 aka Letta / Agent Memory / 长期记忆 2023 Intermediate core
IN PLAIN WORDS
Context windows are finite, so long chats and large documents do not fit. MemGPT treats the context as RAM and a database as disk, and lets the model decide via function calls when to store, retrieve or summarise memories – so an agent can remember old conversations and work over documents larger than its window.
Used in: Chat agents with long-term memory (the precursor of the Letta framework)
TIMELINE · prerequisites → this → what builds on it
01Original paper
MemGPT: Towards LLMs as Operating Systems
Charles Packer, Sarah Wooders, Kevin Lin, et al. · 2023
arXiv ↗ ✓ Verified · arXiv API
02Where it sits
Learn the left column first; the right column builds on this one.
Then you can learn
- Nothing in the atlas builds on it yet
Related:Generative Agents (Memory & Reflection)Multi-Agent Conversation (AutoGen)
Left: prerequisites · right: builds on this · dashed: related. Full map →
03Learning resources
Official docs, explainers, courses and reference code. All links checked.
04Where the course teaches it
For AI / data
- Verification: verified, checked 2026-10-22.
- JSON: /api/learn/algorithms.json (id:
memgpt)