Letta
used in
0
stacks
Stateful agent memory framework (formerly MemGPT); core+archival tiers.
Pioneered the MemGPT pattern: distinguish core (always-in-context) from archival (retrieved-as-needed) memory. Strong fit for long-running agents with rich per-user state.
Strengths
- Mature memory architecture (core/archival/recall tiers)
- Open source with managed option
- Designed for long-lived agents
Tradeoffs
- More setup overhead than simple memory layers