mem0

Self-improving memory layer for AI agents — stores user preferences, conversation history, and learned facts with automatic deduplication and recency scoring.

agent-memory-frameworksRecently releasednpm: mem0ai
79
Hero Score
Popularity
72
Performance
85
Ecosystem
75
Maturity
76
Dev Experience
85
⭐ 57,258 stars⬇ 655.6K downloads/wkFirst release: May 2024Last release: May 2026
Async Support: YesPlugin Extensions: HighSpeed: FastDoc Quality: Very highLearning Curve: Easy

Pros

  • Drop-in `add()` / `search()` API that handles embedding, dedup, and ranking under the hood
  • Works with any LLM and any vector store backend (Qdrant, Chroma, pgvector, etc.)
  • Largest agent-memory community with first-class JS sister package

Cons

  • Managed cloud version costs add up at scale
  • Opinionated extraction pipeline can be hard to debug when memories don't surface as expected
  • Less control than raw vector DB + custom retrieval

Alternatives in agent-memory-frameworks

Compare Python Packages with ease.