Zep & Graphiti
Bi-temporal knowledge graph (event time + ingestion time). Strong on chronological reasoning and contradiction tracking. Graphiti is the open-source core.
At a glance
- Type
- Bi-temporal knowledge graph
- Tier
- T1
- Section
- Dedicated memory layers
- Created
- 2024-08
- Latest release
- v0.29.0 2026-04-27
- License
- Apache-2.0
- GitHub
- 25.7k★ +137/mo Python
- Pricing
- Free + paid
- Funding
- $3M total Seed (additional) · 2024-04
Taxonomy
- storage
- graph
- retrieval
- graph-traversal
- persistence
- long-term
- update
- append-only
- unit
- episode
- governance
- auditable
- conflict
- bi-temporal
When to use
Optimised for: memory operation tracing + drift / poisoning detection
Anti-fit: not for use cases that don't run agent workloads in production
Pros & cons
Pros
Bi-temporal graph captures event time + ingestion time, making contradiction tracking and chronological reasoning correct by construction.
Cons
KG storage is heavier than vector for the same data volume; smaller funding base than Mem0 ($2.3M vs $24M).
Claims & capabilities
Graphiti at 24.3k★. Published counter-analysis disputing Mem0's LOCOMO results. $2.3M raised Apr 2024.
Technical surface
- API surface
- REST, SDK: Python, JS/TS, Go
- Backend storage
- Postgres + Neo4j (Graphiti)
- Deployment
- Managed-only
- Embedding model
- multiple supported
- Multi-tenancy
- Logical namespace per user/session; AWS VPC self-hosted option for full data residency; HIPAA BAA on Enterprise plan
- MCP
- native (first-party) — Graphiti MCP server
- A2A
- not documented publicly
- OpenTelemetry
- not documented publicly
Compare Zep & Graphiti with…
Similar systems
Other dedicated memory layers in the catalog, ranked by inbound references.
- Mem0 T1
Universal memory layer for AI agents. Three concurrent stores (vector + graph + KV); LLM-extracted facts; concurrent retrieval via ThreadPoolExecutor.
- Cognee T1
"Extract–Cognify–Load" pipeline that turns raw input into a typed, queryable knowledge graph for agent recall.
- Hindsight (Vectorize) T1
Standalone memory service from Vectorize. Open source. Biomimetic four-network design (World, Bank, Observation, Opinion). Ships an MCP memory server.
- Memvid T2
Single-file memory layer (one .mv2 file). No DB, no server. Append-only sequence of immutable Smart Frames with timestamps + checksums. Native Rust core (rewritten from Python).
- Supermemory T1
Memory engine with API, app, browser extension, and MCP server. Extracts facts, tracks updates, resolves contradictions, auto-forgets expired info. Plugins for Claude Code, OpenCode, OpenClaw, Hermes.
- AI Singapore SEA-LION T2
SEA-LION-Embedding (March 2026): retrieval + reranking models contrastively trained on 245M text pairs across 10 SE Asian languages. SEA-BED benchmark (169 datasets). SEA-LION v4 (Gemma-based) at 128K context with native function calling.
Related systems
References (1)
- Neo4j depends on at runtime — backend-storage cell: Postgres + Neo4j (Graphiti)
Referenced by (9)
- AutoGen Memory integrates with — Integrates with Mem0/Zep rather than building deep memory natively.
- CrewAI Enterprise depends on at runtime — adjacent-infrastructure cell: OSS CrewAI (already in catalog); typically paired with Mem0 / Zep for memory + Langfuse / LangSmith for observability
- EverMemOS cites — S2 isInfluential citation
- FalkorDB integrates with — MCP server via Graphiti integration
- FalkorDB depends on at runtime — or Mem0; MCP server via Graphiti integration; context graphs for long-term agent memory.
- Graphiti MCP Server (Zep) builds on — Graphiti is the open-source core; MCP exposure of Graphiti's bi-temporal KG
- MAGMA cites — S2 isInfluential citation
- MemR³ cites — S2 isInfluential citation
- RGMem cites — S2 isInfluential citation