What It Is
Graphiti is an open-source memory framework focused on temporally-aware knowledge graphs. It stands out because it is not just another vector-memory wrapper: it is meant for agent systems where facts change over time, relationships matter, and recall quality depends on modeling how context evolves.
Best For
- Developers building graph-shaped memory systems for agents
- Teams that care about changing facts, entity relationships, and temporal context
- Readers comparing managed memory products against lower-level open-source frameworks
Core Use Cases
- Giving agents long-term recall over evolving entities and relationships
- Building knowledge-graph memory for research or multi-step workflows
- Modeling time-sensitive context that simple embedding recall can miss
- Creating custom memory pipelines with more control than managed products provide
Integrations
- Neo4j-backed graph storage
- OpenAI model workflows
- Anthropic model workflows
- Gemini model workflows
- MCP-adjacent tool environments
Deployment
- Local development and experimentation
- Self-hosted setups for teams controlling their own graph stack
Pricing
Graphiti itself is open-source. In practice, the cost profile depends on the graph database, infrastructure, and model providers used in the surrounding stack.
Pros
- Strong differentiation from generic memory layers
- Open-source and flexible for custom graph-based workflows
- Good fit when temporal reasoning matters
Cons
- More technical to adopt than managed memory products
- Requires graph and schema decisions that many teams are not ready for
- Not ideal for readers who only need lightweight memory persistence
Alternatives
- Zep
- Letta
- Mem0
- Custom graph or retrieval pipelines
Related Tools
- Zep
- Letta
- Mem0
- LangGraph
- OpenAI Agents SDK