MemOS vs supermemory
MemTensor's memory OS for LLM agents: one API over graph-structured, multi-modal memory with hybrid retrieval and skill evolution — hosted, self-hosted (Neo4j + Qdrant) or local plugins. — versus — Memory and context engine for AI: fact extraction, user profiles, contradiction handling and forgetting, hybrid RAG + memory search, connectors, agent plugins and a one-binary local mode.
Both are full memory stacks with user-level recall, hybrid retrieval and agent plugins; MemOS emphasizes memory cubes and skill evolution, Supermemory fact extraction, user profiles and connectors.
| MemOS | supermemory | |
|---|---|---|
| Stars | 12k | 31k |
| Forks | 1.1k | 2.7k |
| Language | TypeScript | TypeScript |
| License | Apache-2.0 | MIT |
| Last activity | 5 days ago | today |
| Topics | memory, skills | memory, rag |
| Curated connections | 6 | 4 |
MemOS — the curator's take
Pick MemOS when you want a full memory stack rather than a vector wrapper: graph-structured memories you can inspect and correct in natural language, isolated or shared 'memory cubes' across users and agents, async ingestion, and traces that crystallize into reusable skills. The fastest path is a plugin — local SQLite for Hermes, OpenClaw or DeepSeek Harness, zero infra. Self-hosting the service means running Neo4j and Qdrant plus LLM and embedder keys, which is heavy for a single-user bot. Read the headline LoCoMo/LongMemEval numbers with care: the comparison runs on OmniMemEval, MemTensor's own harness. Skip it if you only need session recall for one coding agent — claude-mem or engrim is far less machinery.
supermemory — the curator's take
The most complete memory product on the map: fact extraction with temporal updates and forgetting, ~50ms user profiles, RAG and memory in one query, connectors for Drive, Gmail, Notion and GitHub, and plugins for Claude Code, Codex, Cursor, OpenCode and Hermes. `npx supermemory local` runs the same Memory API on your machine with local embeddings, so prototyping does not need their cloud; connectors are not in the local feature list, so check before planning around them. Treat the '#1 on every benchmark' banner as vendor-reported. Skip it if you want memory you can read and edit as files (acontext, okf-agent-memory) or only need session recall for one coding agent.