About
Cognee is the long-term memory of the AI stack. Instead of every agent forgetting at the end of a session, important facts are ingested into a knowledge graph and retrieved semantically when needed. It exposes both a REST API and an MCP (Model Context Protocol) endpoint, so any agent — assistant, orchestrator, or worker — can remember and recall with the same interface. The graph approach keeps retrieval explainable: you can trace why a memory surfaced.
What it does
- Knowledge-graph backed long-term memory
- Semantic search and graph retrieval over ingested facts
- REST API plus MCP endpoint for agent access
- Dataset-scoped memories with clean isolation
- Explainable retrieval — traceable memory provenance
Integrations
Stack
- Python
- FastAPI
- PostgreSQL
- MCP
This is an internal endpoint — it powers the platform over a REST API and is not a public UI.