AI Agent Memory Infrastructure
16.8k 2026-04-25
topoteretes/cognee
An open-source knowledge engine designed to build personalized and dynamic memory for AI agents, enabling them to learn and provide relevant context.
Core Features
Unified data ingestion and multimodal knowledge infrastructure (graph/vector search, ontology grounding).
Supports persistent and learning AI agents with context management and cross-agent knowledge sharing.
Ensures reliable and trustworthy agents through user/tenant isolation, traceability, and audit trails.
Combines vector search, graph databases, and cognitive science for meaning-based search and relationship mapping.
Runs locally and ingests data in any format or structure.
Detailed Introduction
Cognee is an open-source knowledge engine that empowers AI agents with dynamic and persistent memory. It unifies data ingestion from various formats and structures, leveraging a combination of vector search, graph databases, and cognitive science to continuously learn and provide context. This enables AI agents to understand relationships, learn from feedback, and share knowledge, fostering more reliable, trustworthy, and personalized interactions. Cognee aims to be the foundational knowledge infrastructure for advanced AI agent development.