Cognee

Cognee is an open-source agent memory platform that captures context, turns it into graph memory, and enables AI agents to recall it across sessions. The platform combines graph and vector databases to uncover hidden relationships and new patterns in data, providing a scalable semantic layer for AI applications powered by large language models and knowledge graphs. Cognee offers first-party integrations with Claude Code, Cursor, Codex, LangGraph, and any MCP-compatible client, allowing agents to read and write memory without custom glue code. The platform supports custom ontologies, custom data models, and permission controls. Key use cases include serving as a second brain for notes and decisions, sales and deal intelligence, technical documentation and manuals, and durable memory for coding agents. Cognee's architecture enables on-premise deployment without requiring graph or vector database configuration, with cited answers that preserve source attribution. The platform processes over 5 million SDK runs per month and has 29,600 GitHub stars. Case studies include Bayer deploying agentic research memory, Knowunity building a POC in two days across 40,000 students, SlideSpeak adding shared context memory for slide creation, and Dynamo enriching data for thousands of customers with personalized support. Cognee offers a free tier with 1 million tokens, a standard tier at $2.50 per million tokens, and enterprise plans with dedicated support and BYO cloud. Founded in 2024 and headquartered in Berlin, Germany, Cognee has raised $9.1M in seed funding from investors including 42 Capital, Angel Invest Ventures, Combination VC, Pebblebed, and Vermilion Cliffs Ventures.