topoteretes/cognee

Cognee is the open-source AI memory platform for agents. Give your AI agents persistent long-term memory with small models for free

Cognee – Open‑source AI memory platform

What it is – Cognee is a Python‑based library (with a CLI, REST API, UI and SDKs for several languages) that lets you turn text, code, documents, tickets, etc. into a knowledge graph that an AI agent can query later. It provides persistent, searchable “long‑term memory” for agents, so they can recall facts or lessons from previous runs without re‑reading the original sources.

Key capabilities

  • Ingestion – cognee.remember extracts entities and relations from raw text (using a tiny local GLiNER model) or from code, stores them as graph nodes, and creates vector embeddings for similarity search.
  • Retrieval – cognee.recall finds relevant chunks, graph sub‑structures or code symbols and can either return the raw source or, if an LLM key is supplied, generate a natural‑language answer.
  • Improvement / feedback – cognee.improve lets you add annotations or corrections that are merged back into the graph, turning session‑specific knowledge into permanent memory.
  • Forget – cognee.forget removes items or whole datasets.
  • LLM‑optional – All core steps work with free local models; an OpenAI (or Ollama, etc.) key can be added later to get answer generation and vision/transcription support.
  • Multi‑language SDKs – Python API, TypeScript package, Rust crate, and a REST endpoint, plus ready‑made plugins for Claude Code, Codex, OpenClaw, etc.
  • Deployability – Run locally via pip/uv, Docker Compose (API + UI + MCP server), or a single‑Postgres stack that combines graph, vector, and relational storage.

Typical use‑cases

  • Building a Company Brain: ingest internal docs, tickets, code, and conversation logs so both humans and agents can query the collective knowledge.
  • Giving autonomous agents memory across sessions, so they can reuse learned rules or past decisions.
  • Grounding agents in a domain‑specific ontology by defining custom entity types and relationships.

How to get started

  1. Install the package (e.g. uv pip install "cognee[gliner]").
  2. Run the quick‑start script or CLI commands to remember a sentence and recall a question – no API key needed.
  3. Optionally set LLM_API_KEY to enable generated answers.
  4. Explore the bundled demo (cognee-cli demo) or launch the UI (cognee-cli -ui).

Ecosystem – The main repo hosts the core engine; a separate cognee‑integrations repo contains plugins for Claude Code, Codex, OpenClaw, and other “memory client protocols” (MCP). Community adapters are listed under cognee‑community.

Research backing – The project is accompanied by a 2025 arXiv paper on linking knowledge graphs with LLMs for complex reasoning, and a BEAM benchmark showing high scores (0.79 for 100 K‑token contexts).

License & community – MIT‑style open source, with a Discord, Reddit community, and contribution guide.

All details are taken directly from the repository’s README.

Related

  • Project
    doobidoo/mcp-memory-service

    A self‑hosted memory backend for LLM agents that provides a REST API, CLI, OAuth, and a web dashboard. It stores facts, decisions, and causal knowledge graphs locally (SQLite + ONNX embeddings) or in hybrid/cloud mode, offering fast (≈5 ms) semantic search, tag‑scoped retrieval, and optional LLM‑generated summaries. Works with LangGraph, CrewAI, AutoGen, Claude Desktop, ChatGPT, and many other tools.

  • Project
    24kchengYe/MemoMind

    MemoMind is a locally hosted, GPU‑accelerated memory system for AI coding agents. It stores extracted facts from chats, documents, and daily life events in PostgreSQL + pgvector, builds a knowledge graph, and offers fast 4‑way hybrid retrieval (semantic, BM25, graph, temporal). The agent can retain new information, recall relevant memories, and reflect across the whole store. All data stays on the user’s machine, works with any OpenAI‑compatible LLM, and includes a web dashboard for browsing and exporting memories.

  • Dispatch
    Bolnee-Chat: Self-Hosted RAG Chatbot for Business Websites

    Bolnee-Chat is an open-source, self-hosted RAG chatbot platform that allows businesses to integrate grounded AI assistants into their websites using a two-line embed snippet.

  • Project
    liliang-cn/cortexdb

    CortexDB is a pure‑Go library that provides a local AI memory store and RDF knowledge graph in a single SQLite (or PostgreSQL) file. It supports vector indexes, hybrid full‑text search, SPARQL/OWL inference, and can be shared via a lightweight gRPC server. Plugins let Claude Code, Codex and other assistants use `/remember` and `/recall` commands, while an optional UI mod adds automatic recall previews and idle‑capture of conversation snippets. Install with `go get`, run the optional `cortexdb-grpc` server for multi‑agent sharing, and configure embeddings via environment variables. MIT‑licensed.

  • Project
    Signet-AI/signetai

    Signet AI is an open‑source service that automatically extracts and structures information from files, chats and web sources into a searchable knowledge graph. Connected AI agents (Claude Code, OpenCode, Gemini CLI, etc.) can query this graph for relevant context before generating responses, preserving memory across model switches. It offers real‑time sync with Obsidian, GitHub, Notion, Discord, a background “dreaming” pipeline, a web/desktop dashboard, encrypted secret storage, and optional telemetry. Installable via a one‑liner script, npm/Bun, or Docker, it runs locally on macOS, Linux and Windows.