# Cognee Cognee is an open-source knowledge engine that provides persistent, structured memory for AI agents. It ingests data in any format, builds a [[Knowledge Graph (KG)]] with vector embeddings, and enables combined semantic/structural search. It goes beyond basic [[Retrieval-Augmented Generation (RAG)]] by grounding data in ontologies, learning from feedback, and self-tuning over time. ## Key Features - **Unified ingestion**: supports 38+ data source types (text, PDF, audio, image, etc.) - **Triple storage architecture**: relational store (provenance/metadata), vector store (semantic embeddings), graph store (entities/relationships) - **Four main operations**: `add` (ingest), `cognify` (build knowledge graph), `memify` (post-processing enrichment via feedback loops), `search` (semantic, structural, or hybrid) - **Ontology grounding**: connects data to external RDF/XML ontologies for domain-specific knowledge - **Self-improvement**: rated responses feed back into edge weights; memory sharpens with use - **Multi-tenant isolation**: agentic user/tenant isolation with traceability - **Cross-agent knowledge sharing**: agents can share and learn from a common memory layer ## Technical Details - Written in Python (3.10-3.13) - Pydantic-based DataPoints as atomic knowledge units (become graph nodes and vector entries simultaneously) - ECL pipeline architecture (Extract, Cognify, Load) with three complementary storage backends - Ships with local-only defaults; swap in production backends (Neo4j, Amazon Neptune, etc.) - Integrations with Claude Agent SDK, OpenAI Agents SDK, LangGraph, Google ADK, n8n - Rust engine upcoming for edge/on-device agents - Apache-2.0 license - 14.7k GitHub stars, 80+ contributors ## Who Created It Founded by Vasilije Markovic (background in big data engineering and cognitive science) in Berlin in 2024. Company: Topoteretes. Raised $7.5M seed round (Feb 2026) led by Pebblebed (co-founded by Pamela Vagata, co-founder of OpenAI). ## References - https://github.com/topoteretes/cognee - https://www.cognee.ai - https://docs.cognee.ai ## Related - [[Knowledge Graph (KG)]] - [[Retrieval-Augmented Generation (RAG)]] - [[RAG Pipelines]] - [[AI Subagents]]