← All comparisons · Compare
Memorg vs Zep
Temporal knowledge-graph memory · cloud + OSS
Zep and Memorg make different bets. Zep organises memory as a temporal knowledge graph and offers a cloud service; Memorg is a local-first, single-file SQLite library with deterministic retrieval and no separate service to run.
| Feature | Memorg | Zep | Advantage |
|---|---|---|---|
| Primary form factor | Python library + MCP server | Service + SDKs (open-source Graphiti core, hosted Zep Cloud) | Comparable |
| Storage model | Hierarchical: session/conversation/topic/exchange in SQLite | Temporal knowledge graph of entities and relationships | Comparable |
| Facts that change over time | Stored exchanges + memory items, with recency in the score | Temporal knowledge graph tracks how facts change over time | Zep |
| Retrieval ranking | Deterministic blend: semantic + recency + importance | Search and traversal over the knowledge graph | Memorg |
| Vector index | USearch, in-file alongside SQLite | Managed as part of the service | Comparable |
| Separate infrastructure | None — no separate vector DB or cloud service | Runs as a service (graph + storage) | Memorg |
| Default deployment | Local — one process, one file | Hosted (Zep Cloud) or self-host the open-source core | Comparable |
| Token-budget aware | Yes — fits results to a session-level token budget | Configurable on the client | Comparable |
| MCP server | Ships with the package (memorg-mcp / FastMCP) | Via integrations / not the primary interface | Memorg |
| Maturity | Alpha on PyPI, MIT | Established project with production users | Zep |
| License | MIT | Open-source core (Graphiti) + commercial cloud | Comparable |
Zep rows describe its publicly documented shape — a temporal knowledge graph (Graphiti) with a hosted cloud offering; specifics vary by version and deployment. Memorg rows are grounded in the Memorg README and docs. This comparison publishes no benchmark numbers.
Pick Memorg when
- ▸You want a small, deterministic library that runs in your process with no service to operate
- ▸You want a single SQLite file you can copy, version, and back up
- ▸You do not need entity-relationship (graph) recall as the dominant query shape
- ▸You want a first-party MCP server and a framework-agnostic drop-in
Pick Zep when
- ▸Your data is full of facts that change over time and you want a temporal knowledge graph to track them
- ▸You want a hosted offering (Zep Cloud) with a control plane
- ▸Entity-relationship recall across a graph is central to your product
- ▸You prefer a managed service over running the memory layer in-process
Still deciding?
Most teams end up running both kinds of system in different parts of their stack. Try Memorg where the answer is "I just need recall over a session," and keep Zep where you need the rest of its surface.