Definition
Plain language
A system that gives an AI assistant a managed memory, swapping information in and out the way a computer juggles what it keeps in active memory.
As stated in the literature
An LLM memory-management framework that treats the context window like RAM and external storage like disk, paging information in and out to extend effective memory; cited as a precursor to OS-inspired LLM serving systems.
Why it matters: It lets an assistant act as if it has a much larger memory than its fixed context window allows, so long conversations don't simply fall off the edge and vanish.
For example, when a chat assistant runs out of room to hold everything you've discussed, MemGPT tucks older details into storage and pulls them back when they become relevant again.
Heard on the show
“MemGPT and AIOS are kindred-spirit systems in that lineage.”Episode 016 — Why Your Coding Agent Stalls While the GPU Runs Hot