Cognee

What is Cognee about?

Cognee enables AI agents to store, retrieve, and reason over past interactions, documents, images, and audio transcriptions. It replaces traditional Retrieval‑Augmented Generation (RAG) workflows with a unified, low‑code solution that builds and queries knowledge graphs and vector databases.

How to use Cognee?

1. Install the package (pip, poetry, or uv). Example with pip:

pip install cognee

2. Configure your LLM API key, e.g., via environment variable or a .env file:

import os
os.environ["LLM_API_KEY"] = "YOUR_OPENAI_API_KEY"

3. Add data and run the default pipeline:

import cognee, asyncio

async def main():
    await cognee.add("Natural language processing (NLP) is ...")
    await cognee.cognify()
    results = await cognee.search("Tell me about NLP")
    for r in results:
        print(r)

asyncio.run(main())

4. Optionally launch the Cognee UI for visual ingestion and querying.

Key Features

• Interconnect and retrieve past conversations, documents, images, and audio transcriptions.
• Fully replaces RAG systems, reducing developer effort and cost.
• Load data into graph (e.g., Neo4j) and vector databases using only Pydantic models.
• Ingest from 30+ sources (files, web pages, APIs, etc.).
• Supports Python 3.10‑3.13; installable via pip, poetry, or UV.
• Hosted "Cogwit" beta offering a managed AI memory service.
• Built‑in UI for visual data cognition and query.

Use Cases

• AI assistants that remember user preferences and conversation context.
• Knowledge‑graph construction for enterprise data and research labs.
• Multimodal retrieval, enabling agents to search across text, images, and audio.
• RAG replacement in LLM‑driven applications, cutting latency and API costs.
• Rapid prototyping of memory‑enabled agents for hackathons or proof‑of‑concepts.

FAQ

Q: Which LLM providers are supported? A: Any provider that offers an API key compatible with the OpenAI request schema; local models via Ollama are also supported.

Q: Do I need a vector database separate from the graph store? A: Cognee can sync embeddings to both a graph database (e.g., Neo4j) and a vector store; you can choose one or both.

Q: How do I set environment variables without editing code? A: Use a .env file based on the provided .env.template or export variables in your shell.

Q: Is there a hosted version? A: Yes, the Cogwit beta offers a fully‑hosted AI memory service at https://platform.cognee.ai/.

Q: Can I extend the ingestion pipeline? A: The ECL architecture is modular; you can add custom extractors, transformers, or loaders via Python plugins.

Cognee - Accurate and Persistent AI Memory

Use your data to build personalized and dynamic memory for AI Agents. Cognee lets you replace RAG with scalable and modular ECL (Extract, Cognify, Load) pipelines.

About Cognee

Cognee is an open-source tool and platform that transforms your raw data into persistent and dynamic AI memory for Agents. It combines vector search with graph databases to make your documents both searchable by meaning and connected by relationships.

You can use Cognee in two ways:

1. Self-host Cognee Open Source, which stores all data locally by default.
2. Connect to Cognee Cloud, and get the same OSS stack on managed infrastructure for easier development and productionization.

Cognee Open Source (self-hosted):

• Interconnects any type of data — including past conversations, files, images, and audio transcriptions
• Replaces traditional RAG systems with a unified memory layer built on graphs and vectors
• Reduces developer effort and infrastructure cost while improving quality and precision
• Provides Pythonic data pipelines for ingestion from 30+ data sources
• Offers high customizability through user-defined tasks, modular pipelines, and built-in search endpoints

Cognee Cloud (managed):

• Hosted web UI dashboard
• Automatic version updates
• Resource usage analytics
• GDPR compliant, enterprise-grade security

Basic Usage & Feature Guide

To learn more, check out this short, end-to-end Colab walkthrough of Cognee's core features.

Quickstart

Let’s try Cognee in just a few lines of code. For detailed setup and configuration, see the Cognee Docs.

Prerequisites

• Python 3.10 to 3.13

Step 1: Install Cognee

You can install Cognee with pip, poetry, uv, or your preferred Python package manager.

uv pip install cognee

Step 2: Configure the LLM

import os
os.environ["LLM_API_KEY"] = "YOUR OPENAI_API_KEY"

Alternatively, create a .env file using our template.

To integrate other LLM providers, see our LLM Provider Documentation.

Step 3: Run the Pipeline

Cognee will take your documents, generate a knowledge graph from them and then query the graph based on combined relationships.

Now, run a minimal pipeline:

import cognee
import asyncio


async def main():
    # Add text to cognee
    await cognee.add("Cognee turns documents into AI memory.")

    # Generate the knowledge graph
    await cognee.cognify()

    # Add memory algorithms to the graph
    await cognee.memify()

    # Query the knowledge graph
    results = await cognee.search("What does Cognee do?")

    # Display the results
    for result in results:
        print(result)


if __name__ == '__main__':
    asyncio.run(main())

As you can see, the output is generated from the document we previously stored in Cognee:

  Cognee turns documents into AI memory.

Use the Cognee CLI

As an alternative, you can get started with these essential commands:

cognee-cli add "Cognee turns documents into AI memory."

cognee-cli cognify

cognee-cli search "What does Cognee do?"
cognee-cli delete --all

To open the local UI, run:

cognee-cli -ui

Demos & Examples

See Cognee in action:

Persistent Agent Memory

Cognee Memory for LangGraph Agents

Simple GraphRAG

Watch Demo

Cognee with Ollama

Watch Demo

Community & Support

Contributing

We welcome contributions from the community! Your input helps make Cognee better for everyone. See CONTRIBUTING.md to get started.

Code of Conduct

We're committed to fostering an inclusive and respectful community. Read our Code of Conduct for guidelines.

Research & Citation

We recently published a research paper on optimizing knowledge graphs for LLM reasoning:

@misc{markovic2025optimizinginterfaceknowledgegraphs,
      title={Optimizing the Interface Between Knowledge Graphs and LLMs for Complex Reasoning},
      author={Vasilije Markovic and Lazar Obradovic and Laszlo Hajdu and Jovan Pavlovic},
      year={2025},
      eprint={2505.24478},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={https://arxiv.org/abs/2505.24478},
}