cognee review

Q: How do I install cognee?

Run uv pip install "cognee[gliner]". The cognee package needs Python 3.10 or newer, below 3.15. The install section above has the other options from the project's README.

Cognee is an open-source Python memory platform that turns documents, code, and conversations into a self-hosted knowledge graph AI agents can search.

Updated Nov 1, 2026. Review written Oct 22, 2026 from the project's own sources; usage data refreshed every Monday.

TL;DR

Cognee gives AI agents persistent long-term memory across sessions by building a knowledge graph from documents, code, and conversations. It is for developers building agents or shared team knowledge bases. Cognee can run locally on small CPU models with no API key, and it combines graph, vector, and relational retrieval.

33kGitHub stars
+265new stars, week of Oct 25
24.9kdownloads a week, −1% vs 4-week avg
#24in RAG, data and memory this week

cognee at a glance

Category: RAG, data and memory
Maintained by: Topoteretes (organization, @topoteretes on GitHub)
License: Apache 2.0
Language: Python, TypeScript, JavaScript
Repository: github.com/topoteretes/cognee
Website: www.cognee.ai
PyPI package: cognee 1.6.5
Latest release: v1.6.5 on GitHub, Oct 16, 2026
First published: Aug 16, 2023

What cognee does

Cognee converts text into entities, relationships, and searchable chunks, and converts code into a graph of symbols and dependencies.
Cognee provides remember, recall, improve, and forget operations to store, query, enrich, and delete memory.
Cognee distills accepted lessons from agent sessions into permanent memory that later sessions can retrieve.
Cognee supports custom data models and ontologies to structure memory around the entities your application needs.
Cognee offers plugins for Claude Code, Codex, and OpenClaw, plus an MCP server for Cursor, Cline, and other MCP clients.
Cognee can import memory from Mem0, Letta, Zep, or Graphiti using the COGX exchange format.
Cognee bundles Google Drive and Gmail OAuth connectors in the SDK and supports Slack conversation import and sync.

How it works

Cognee runs as a Python library, a CLI, or a Docker API server. With no LLM key, Cognee uses a local GLiNER model and a local embedding model. Adding a key enables generated answers through OpenAI, other providers, or Ollama. Agents connect through plugins, MCP, Python, TypeScript, Rust, or REST. The memory layer can run on a single Postgres instance as a demo feature.

How to install and start using cognee

With uv (includes local extraction model)

uv pip install "cognee[gliner]"

Claude Code plugin

claude plugin install cognee-memory@cognee

Run API with Docker Compose from source

docker compose --profile ui --profile mcp up

Requirements

Python 3.10 to 3.14
An LLM API key is optional and is needed only for generated answers and media processing
Node.js/npm for the local UI launcher, and Docker for its MCP service

First steps

Install Cognee with the gliner extra so it can extract graphs locally without an LLM key.
Store a fact with cognee-cli remember "Marie Curie was born in Warsaw." -d local_quickstart.
Query it with cognee-cli recall "Where was Marie Curie born?" -d local_quickstart.
Set LLM_API_KEY to get generated answers instead of retrieved passages.

Who cognee is for

Agent builders who want memory that persists across runs and sessions
Teams building a shared company knowledge base from docs, tickets, code, and conversations
Developers who want self-hosted graph memory that runs locally without an API key
Applications that need memory structured by domain-specific ontologies

Limitations to know

Cognee's bundled GLiNER extractor is a demo; the README directs users to contact the team for a higher-accuracy production version.
Cognee's Postgres graph store is a demo feature; the production-ready version is a licensed product.
Cognee needs LLM configuration for generated answers and for media requiring vision or transcription models.
Cognee's Docker API defaults to multi-tenant mode and requires authentication unless access control is disabled.

Is cognee growing? Usage data

In the week of Oct 25–31, 2026, cognee ranked #228 on the Devtoolwatch Board and #24 in RAG, data and memory, with a momentum score of 40 out of 100. Its main package had 24.9k downloads that week, −1% against its own 4-week average.

New GitHub stars per week for the last 12 weeks, from GitHub's stargazer history (today's stargazers by the week they starred).

Show the weekly numbers

New GitHub stars per week for cognee
Week of	New stars
Aug 9, 2026	224
Aug 16, 2026	187
Aug 23, 2026	189
Aug 30, 2026	183
Sep 6, 2026	197
Sep 13, 2026	217
Sep 20, 2026	179
Sep 27, 2026	378
Oct 4, 2026	665
Oct 11, 2026	511
Oct 18, 2026	330
Oct 25, 2026	265

On Hacker News

Show HN: Cognee – Open-Source AI Memory Layer That Remembers Context, 9 points and 2 comments (Jun 3, 2025)
Show HN: Auto-optimizing deterministic LLM outputs using knowledge graphs, 7 points and 0 comments (Apr 22, 2024)
Show HN: Cognee – Turn RAG and GraphRAG into custom dynamic semantic memory, 6 points and 1 comments (Feb 13, 2025)

The Board ranks tools by four signals each week: new stars, new stars for the repo's size, added downloads and download growth beyond the market. How the ranking works.

cognee alternatives

Other RAG, data and memory tools on the Board in the week of Oct 25–31, 2026, by momentum:

RAG, data and memory ranked by momentum, the week of Oct 25–31, 2026
Rank	Tool	Stars	Downloads/wk
#1	XERJXERJ is a local-first search engine and MCP server for AI agents, packaged as one Rust binary with an Elasticsearch-compatible API.	17.9k	–
#2	MemvidMemvid is a single-file memory layer for AI agents that packages data, embeddings, search indexes, and metadata into one portable .mv2 file.	17.3k	–
#3	LEANNLEANN is a Python vector database and RAG library that recomputes embeddings on demand to keep indexes small on personal devices.	13.7k	–
#4	GraphifyGraphify is an AI coding assistant skill that turns a codebase, with its docs, PDFs, images and videos, into a queryable knowledge graph.	140k	480.7k
#5	MarkItDownMarkItDown is a lightweight Python utility from Microsoft that converts PDFs, Office documents, images, audio and other files to…	193.2k	4.7M
#6	FirecrawlFirecrawl is an API and open source project to search, scrape, crawl, and interact with web pages, returning LLM-ready markdown or JSON…	197.4k	2.9M

All RAG, data and memory tools, ranked

Recent cognee releases

v1.6.5, Oct 16, 2026
v1.6.5.dev0, Oct 16, 2026
v1.6.4, Oct 13, 2026
v1.6.4.dev0, Oct 13, 2026
v1.6.3, Oct 7, 2026

cognee FAQ

Is cognee free and open source?

Yes. cognee's source code is on GitHub (topoteretes/cognee) under the Apache 2.0 license, so you can use it for free. Its website also links to a pricing page (https://www.cognee.ai/pricing) for any paid plans.

How do I install cognee?

Run uv pip install "cognee[gliner]". The cognee package needs Python 3.10 or newer, below 3.15. The install section above has the other options from the project's README.

Is cognee still maintained?

Yes. The topoteretes/cognee repository was last updated on Oct 21, 2026, and the latest release, v1.6.5 on GitHub, came out on Oct 16, 2026.

How popular is cognee?

cognee has 33k GitHub stars as of Nov 1, 2026. It gained 265 new stars in the week of Oct 25–31, 2026, against a 12-week average of 294 a week. Its main package had 24.9k downloads that week, −1% against its own 4-week average. That made it #24 among RAG, data and memory tools on the Devtoolwatch Board.

cognee vs Mem0: how do they compare?

cognee and Mem0 are both tracked by Devtoolwatch as RAG, data and memory tools. cognee has 33k GitHub stars and Mem0 has 68.2k. In the week of Oct 25–31, 2026, cognee's main package had 24.9k downloads and Mem0's had 884.1k. Mem0 is a memory layer for AI agents and apps that stores and retrieves user, session, and agent context across interactions. Mem0 review.

cognee vs Graphiti: how do they compare?

cognee and Graphiti are both tracked by Devtoolwatch as RAG, data and memory tools. cognee has 33k GitHub stars and Graphiti has 32.4k. In the week of Oct 25–31, 2026, cognee's main package had 24.9k downloads and Graphiti's had 211k. Graphiti is an open-source Python framework for building and querying temporal knowledge graphs, called context graphs, for AI agents. Graphiti review.

What are the best alternatives to cognee?

Other RAG, data and memory tools on the Devtoolwatch Board in the week of Oct 25–31, 2026, by momentum: XERJ (#1), Memvid (#2), LEANN (#3), Graphify (#4), MarkItDown (#5). The alternatives table above has their numbers.

Is cognee open source?

Yes. Cognee is open source under the Apache-2.0 license, and its code is on GitHub at topoteretes/cognee.

How does cognee work?

Cognee turns text into entities, relationships, and searchable chunks, and turns code into a graph of symbols and dependencies. At query time, Cognee selects relevant graph, vector, or code context. Your application can inspect that evidence and use it to answer a question or continue an agent task.

Does cognee need an LLM API key?

No. Cognee runs locally with small CPU models and needs no API key. Text ingestion, retrieval, and session storage work without an LLM. An API key is required only for generated answers and for media processing.

Which agents can use cognee memory?

Cognee has plugins for Claude Code, Codex, and OpenClaw. Cursor, Cline, and other MCP clients connect through the Cognee MCP server. Applications can also use the Python API, the TypeScript SDK, Cognee-RS for Rust, or the REST API.

Can cognee import memory from other systems?

Yes. Cognee can import memory from Mem0, Letta, Zep, or Graphiti using the COGX exchange format.

Sources

cognee README badge

Maintainers can show the tool's weekly Board rank in their README:

[![Devtoolwatch](https://index.devtoolwatch.com/badge/cognee.svg)](https://index.devtoolwatch.com/tools/cognee/)

How this page is made: Devtoolwatch writes each review from the project's README, releases, package metadata and website, with an AI model, then checks every claim against the source it cites and drops what the sources don't support. The numbers come from Devtoolwatch's daily GitHub snapshots, GitHub's stargazer history, and the npm and PyPI download APIs. Spotted an error? Email [email protected].