cognee

Tracked

Open-source AI memory platform that gives agents persistent long-term memory across sessions, backed by a self-hosted knowledge graph engine.

Author topoteretes Open Sourced 2023-08-16 Last Commit Unknown

Introduction

cognee is an open-source memory platform for AI agents that delivers persistent long-term memory across sessions and tools. It ingests unstructured data, models it into a self-hosted knowledge graph, and exposes a retrieval layer that keeps agent context coherent and grounded over time.

Key Features

  • Persistent long-term memory that survives across sessions and agentic runs
  • Self-hosted knowledge graph engine with vector and graph retrieval
  • Pluggable LLM, embedding, and vector/graph store backends
  • GraphRAG-style ingestion that turns documents into connected knowledge
  • Framework-agnostic integration with popular agent stacks

Use Cases

  • Giving custom agents durable memory without vendor lock-in
  • Building GraphRAG pipelines over enterprise documents
  • Maintaining coherent context for long-running, multi-session agents
  • Self-hosting memory and knowledge for data-sensitive workloads

Technical Highlights

  • Combines vector databases and knowledge graphs in one memory engine
  • Data-point centric architecture that separates ingest, processing, and retrieval
  • Runs locally or in private infrastructure with no mandatory cloud dependency
  • Apache-2.0 licensed with a modular, extensible backend layer