Mem0 vs Cognee
Two of the top projects in Memory, side by side: who leads on each score part and who each one suits, then setup, license, activity and what each review found.
Mem0
Long-term memory layer that stores and retrieves user facts for LLM agents
Cognee
Memory engine that turns documents and code into a knowledge graph
Verdict
Mem0 ranks higher, 89 to 85. Of the 4 score parts we measure for both, Mem0 leads on 1, Cognee on 1 and 2 are even.
- AdoptionMem0 leads by 19 points, 97 to 78: a wide gap.
- FreshnessEven at 100.
- MaintenanceEven at 76.
- Agent-readyCognee leads by 15 points, 85 to 70: a wide gap.
Pick Mem0 if you are
Developers adding persistent per-user memory to agents or chatbots
- Python (pip) and npm SDKs plus a CLI, so it embeds in existing agents
- Hybrid retrieval: semantic, BM25 keyword and entity linking fused in one search
Pick Cognee if you are
Developers building graph-backed memory for agents and company knowledge
- Keyless mode: local GLiNER extraction and embeddings, no cloud LLM required
- Claude Code and Codex plugins, OpenClaw plugin, MCP server, Python, TypeScript and Rust SDKs
| What we compare | Mem0 | Cognee |
|---|---|---|
| Score parts, out of 100 | ||
| Adoption | 97, widely used | 78, popular |
| Freshness | 100, active | 100, active |
| Maintenance | 76, fair | 76, fair |
| Agent-ready | 70, partly | 85, ready |
| Facts from GitHub and the README | ||
| Stars | 67.7k | 32.7k |
| License | Apache-2.0 (permissive) | Apache-2.0 (permissive) |
| Last commit | Oct 2026 | Oct 2026 |
| Last release | Sep 2026 | Oct 2026 |
| Language | Python | Python |
| Docker | No | Yes |
| GPU | Optional | Not needed |
| arm64 or Apple Silicon | Not stated | Mentioned |
Mem0
Mem0 extracts facts from conversations with an LLM, stores them per user, session or agent, and retrieves them with combined semantic, BM25 keyword and entity matching. It ships as a Python and npm library, a self-hosted server with a dashboard and API keys, a CLI, and a managed cloud platform. The published benchmark numbers come from the managed platform, not the open-source SDK.
Who it is for: Developers adding persistent per-user memory to agents or chatbots
Strengths
- Python (pip) and npm SDKs plus a CLI, so it embeds in existing agents
- Hybrid retrieval: semantic, BM25 keyword and entity linking fused in one search
- Evaluation framework is open-sourced so benchmark numbers can be reproduced
- Supports several LLMs and embedders; defaults are configurable
Weaknesses
- Benchmarks reflect the managed platform; OSS results may differ
- Needs an LLM; default is OpenAI gpt-5-mini and text-embedding-3-small
- BM25 and entity features need the optional mem0ai[nlp] extra and a spaCy model
- Self-hosted dashboard is listed as limited; advanced features are cloud-only
- GPU optional
- Needs LLM provider (default OpenAI), Embedding model (default OpenAI text-embedding-3-small), Docker Compose (self-hosted server), spaCy en_core_web_sm (optional NLP extra)
- Models: OpenAI gpt-5-mini (default), text-embedding-3-small (default), other LLMs and embedders via configuration
- port 3000
Cognee
Cognee builds persistent agent memory by extracting entities, relationships and chunks from text, code and sessions into a graph and vector index with hybrid recall. Without an LLM key it uses local GLiNER extraction and embeddings; adding a key enables generated answers via OpenAI, Ollama or other providers. It runs as a library, CLI, REST API (port 8000), UI (3000) and MCP server (8001).
Who it is for: Developers building graph-backed memory for agents and company knowledge
Strengths
- Keyless mode: local GLiNER extraction and embeddings, no cloud LLM required
- Claude Code and Codex plugins, OpenClaw plugin, MCP server, Python, TypeScript and Rust SDKs
- Imports memory from Mem0, Letta, Zep or Graphiti via the COGX format
- Apache-2.0; research paper and BEAM evaluation published
Weaknesses
- Single-Postgres graph store is a demo; production version is a licensed product
- API defaults to multi-tenant mode; local use needs ENABLE_BACKEND_ACCESS_CONTROL=false
- Bundled GLiNER extractor is a demo; higher-accuracy version requires contacting the vendor
- UI launcher needs Node.js/npm and Docker for its MCP service
- no GPU
- Docker + Compose
- Compose runs Neo4j, PostgreSQL, Redis
- Models: local GLiNER + embeddings (keyless), OpenAI, Ollama, other providers per docs
- port 8000
More in Memory
Building with these?
We help teams pick an open-source AI stack and run it in production. Client work never changes a rank.