mem0
by mem0
Memory layer for AI apps. Personalisation, continuity, and recall as a service.
OSS
mem0
Added 17 May 2026
Overview
mem0 is a memory layer for LLM apps that handles extraction, storage, and recall of user-specific facts across sessions. Drop in front of any LLM call to give an app persistent personalisation. Available as a hosted API or fully self-hosted from the open-source repo.
Best for
Best for
Product teams who want memory without building the storage layer
Use cases
- Personalise an assistant across conversations without rolling your own memory layer
- Persist user preferences, history, and context across sessions
- Add memory to an existing LLM app in a couple of hours
- Audit what the agent remembers about a given user
Notes
Why it matters
Memory is the next layer in LLM apps after retrieval. mem0 packages it as a service you can plug in, instead of a problem you have to architect.
How teams use it in production
Wire mem0 in front of every chat call. Tag memories with user ID. Review the memory store periodically and prune stale facts.
What to watch
The memory layer is becoming the personalisation layer. The vendor who owns that surface for AI apps is a real position to compete for.
Pros
- Hosted or self-hosted, the choice is real
- Drop-in front of any LLM call, minimal refactor
- Memory extraction is opinionated and works well by default
- Audit tools for inspecting what was remembered
Cons
- Hosted version is a SaaS dependency
- Less low-level control than Letta on memory schema
- Overhead for short-lived agents