Cognee

Cognee is an AI memory platform open-sourced by the topoteretes team. The README calls it "the free open-source AI memory platform for agents." You give it documents, code and chat logs, and it organises them into a knowledge graph that your agent can search later, with memory that keeps building up. It's a Python library that plugs into your agent, with optional MCP support and data connectors. What sets it apart is that it can run entirely on local models: there's a path for small local models on a CPU, and it also supports cloud models.

Who it's for:Developers building AI agents that need to remember material over the long term. Readers who don't write code won't get much use from it.

Before you start: You need Python 3.10 to 3.14, installed with pip or uv. The local-model route downloads a model the first time. If you switch to a cloud model, your content goes to that model provider. Google Drive, Gmail and similar connectors need OAuth authorisation and are optional, and hooks and plugins widen what it can touch. Start with a local model and data that isn't sensitive, and keep storage and services from being exposed. The code is free under the Apache-2.0 license; cloud models may cost extra.

Based on the project's public documentation. Not tested by AI Inspiration World.