letta-ai/letta
A platform for stateful agents with self-editing memory, descended from the MemGPT research line: agents manage their own memory blocks inside a bounded context. Agentiquette scores it 77/100 (A-tier), methodology v1.0, scored 2026-07-07.
Key facts
| Category | Memory system |
|---|---|
| License | Apache-2.0 |
| Maintainer | Letta |
| Compatible agents | generic |
| File patterns | |
| Triggering | manual-reference |
| GitHub metrics | 23,689 stars · 2,513 forks · 48 open issues (fetched 2026-07-07) |
| Last commit | 2026-07-03 |
| Created | 2023-10-11 |
Editorial analysis
The intellectual ancestor of half the memory patterns in this index: the MemGPT insight that agents can manage memory as a resource under their own control. As a product it is heavier than file-based patterns and earns that weight only when agents genuinely live for weeks. The self-editing design is powerful and is also exactly the property our Memory Safety Model flags: agent-written history needs an independent audit trail beside it.
Risks
Self-editing memory means the agent curates its own record; pair with external logging if you need ground truth about what actually happened.
Evidence notes: Reliability: memory operations are structured and observable. Eval readiness: research lineage means published evaluation methods exist. Composability: platform-shaped; integrating pieces without the whole takes work. Safety: agent self-editing is by design, mitigations documented.
Install and first run
pip install letta # server: docker run lettaai/lettaAlternatives in this category
A universal memory layer for AI agents: extraction, storage, and retrieval of user and session memories with SDKs across the major agent…
Local-first agent memory over plain Markdown files with an MCP interface: knowledge persists as human-readable notes the agent reads and…
A temporal knowledge-graph memory layer for agents: facts extracted from conversations with validity intervals, so recall can answer what…
Embed this badge
Maintainers of scored entries can embed their live badge; it updates when the score does and links back to this page.
[](https://agentiquette.com/index/repos/letta)Frequently asked questions
Is letta-ai/letta good?
letta-ai/letta scores 77/100 (A-tier) on Agentiquette's methodology v1.0, scored 2026-07-07. Strengths and weaknesses are in the editorial analysis; the score is explained dimension by dimension in the scorecard.
What is letta-ai/letta for?
A platform for stateful agents with self-editing memory, descended from the MemGPT research line: agents manage their own memory blocks inside a bounded context.
Is letta-ai/letta maintained?
Last commit 2026-07-03, 48 open issues at fetch time (2026-07-07). Maintenance is scored at 86 of 100.
Maintainer of letta-ai/letta? Claim this listing and suggest corrections.