cognee: an open-source AI memory platform that provides agents with persistent long-term memory via self-hosted knowledge graphs

What it solves

Cognee provides AI agents with persistent, long-term memory across sessions. It solves the problem of agents forgetting context or failing to connect disparate pieces of information by creating a self-hosted knowledge graph from ingested data, allowing agents to recall and connect information based on both meaning and relationships.

How it works

Cognee transforms raw data into a structured memory layer. It combines vector embeddings for semantic search and graph reasoning for relationship tracking, using cognitive-science-grounded ontology generation to organize knowledge. The platform supports a flexible backend architecture where the entire memory layer—including relationships, embeddings, and session caches—can be run on a single Postgres instance using pgvector, or swapped for dedicated databases like Neo4j or Redis.

Who it’s for

Developers building AI agents that require domain-specific knowledge, long-term persistence, and the ability to learn from feedback and user interactions across multiple sessions.

Highlights

• Unified Memory Layer: Combines vector and graph search into a single infrastructure.
• Cros-Agent Knowledge Sharing: Enables multiple agents to share a persistent knowledge base.
• Flexible Deployment: Supports local development (SQLite/LanceDB), self-hosting via Docker, or a managed cloud service.
• Multi-Language Support: Official clients available for Python, Rust, and TypeScript.
• Agent Integrations: Includes a dedicated plugin for Claude Code to capture and sync session memory.