
Cognee
Open-source memory layer that turns docs and chats into agent-searchable knowledge graphs.
About
Cognee is an open-source AI memory platform: feed it documents, code, tickets, and conversations and it builds a persistent knowledge graph your agents can search and reuse across sessions. It is designed to run locally for free — the default pipeline uses small models on CPU with no API key required — while letting you plug in stronger hosted models. The team frames the flagship use case as a 'Company Brain' shared between humans and agents, and published a research paper on the knowledge-graph/LLM interface in 2025.
What it can do
- **Persistent memory**: documents, code, and conversations become a queryable knowledge graph
- **Local-first**: free CPU pipeline with no API key; bring your own models when ready
- **Agent integration**: designed as the memory backend for agent frameworks
- **Research-backed**: arXiv paper on KG/LLM interfaces (Markovic et al., 2025)
- **Community plugins**: cognee-community repo for add-ons
Setup
```bash
pip install cognee
```
```python
import cognee
# Add documents to memory
await cognee.add("The project deadline moved to Friday.")
await cognee.cognify()
# Ask across everything remembered
answer = await cognee.search("When is the deadline?")
```
Runs locally by default (no API key). Configure hosted LLMs/embeddings via env vars for higher quality. Docs: docs.cognee.ai.
Prerequisites
Python 3.10+. Optional: LLM/embedding provider API keys for higher-quality pipelines (local CPU models work out of the box).
Use cases
Works with
Trust score — 9/10
Scored 9/10 (verified tier, metrics-v1). Evidence: 32,105 GitHub stars (log-scaled vs catalog peers); blended with ~86,048 monthly package downloads; last push 0d ago; 89 releases/12mo; 680 open issues. Manual review: 32k stars / 86k PyPI-mo; free local CPU pipeline, no API key; research paper published; graph quality depends on plugged-in models. License: Apache-2.0.
★ 32,105 GitHub stars
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 8.1/10
- Maintenance · recency, releases, issue load
- 9.2/10
- Openness · license + docs
- 10.0/10
- Manual review · sanity read + red-flag check
- 9.0/10
- · 32,105 GitHub stars (log-scaled vs catalog peers)
- · blended with ~86,048 monthly package downloads
- · last push 0d ago; 89 releases/12mo; 680 open issues
- · manual review: 32k stars / 86k PyPI-mo; free local CPU pipeline, no API key; research paper published; graph quality depends on plugged-in models
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.