Mem0
Universal memory layer for AI agents — multi-signal retrieval, temporal reasoning, 57k+ GitHub stars
Give your AI apps persistent memory and intelligence
Supermemory
Memory layer for AI agents — hybrid vector search, ontology-aware graph, and sub-100ms retrieval
Memory layer for AI agents — hybrid vector search with ontology-aware graph and sub-100ms retrieval. Self-hostable alternative to Pinecone and Mem0.
Compare Letta alternatives
| Tool | Stars | Forks | License | Self-hosted |
|---|---|---|---|---|
| Mem0 | 57.3K | 6.5K | Apache-2.0 | Yes |
| Supermemory | 23.6K | 2.1K | MIT | Yes |
Letta (formerly MemGPT) is a framework for building stateful AI agents with self-editing memory — the agent manages its own context using an OS-inspired hierarchy of in-context and archival memory. The core framework is open source, but Letta also runs a managed cloud platform for hosting and monitoring agents, which is the paid, closed part of the offering.
Mem0 and Supermemory take a different architectural approach — an external memory layer your application calls into, rather than memory the agent edits itself — which is worth comparing if Letta's self-editing model isn't the fit you need.
FAQ
Is there a free, open-source alternative to Letta?
Letta's core framework is itself already open source. Mem0 (Apache-2.0) and Supermemory (MIT) are open-source alternatives if you want an external memory layer instead of Letta's self-editing agent memory model.
What's the best open-source alternative to Letta?
Mem0 is the better pick for a general-purpose external memory layer with multi-signal retrieval. Supermemory is the better pick if low-latency retrieval and graph-based memory structure are the priority.
What are the top open-source alternatives to Letta?
Browse all open-source alternatives to Letta above, or explore the full AI & LLM category for more options.