Graphiti
Temporal knowledge-graph memory for AI agents — extracts entities and relationships from conversations into a Neo4j-backed graph that tracks how facts evolve over time.
agent-memory-frameworksRecently released
60
Hero Score
Popularity
64
Performance
70
Ecosystem
50
Maturity
61
Dev Experience
57
⭐ 26,828 stars⬇ 129.3K downloads/wkFirst release: Aug 2024Last release: May 2026
Async Support: YesPlugin Extensions: MediumSpeed: MediumDoc Quality: HighLearning Curve: Medium
Pros
- • Bi-temporal graph model tracks both event time and recorded time, so superseded facts don't pollute retrieval
- • Built specifically for evolving agent memory rather than static document RAG
- • Powers Zep's hosted memory product, so it's battle-tested at production scale
Cons
- • Requires a Neo4j or FalkorDB backend — heavier infra than vector-only memory
- • LLM extraction calls per ingestion make it slower and pricier than naive embedding writes
- • Graph design and entity schemas take effort to tune for your domain