Awesome AI Memory

Add cognee to your AI App

🌟

A curated list of tools, frameworks, optimizers, and storage layers that give AI apps and agents memory.

Check out Cognee – A semantic memory for AI Apps and Agents

• Browse examples in the Cognee QuickStart
• Contact us at info@topoteretes.com or join us on Discord
• Follow us on Twitter

Welcome to our curated list of AI memory tools

You can see the projects split across various dimensions

1. Open-source projects vs Closed-source projects and companies
2. Graph, Vector, or Both
3. Memory, Framework, Optimizer, or Storage

There might be many more companies we are not aware of. Your feedback and contributions are appreciated! ❤️

Have anything to add? Please open a pull request

Check out Cognee – A semantic memory for AI Apps and Agents

• Browse examples in the Cognee QuickStart
• Contact us at info@topoteretes.com or join us on Discord
• Follow us on Twitter

Contents

• How to use this list
• Types of AI memory
• Memory tools
• LLM frameworks
• Optimizers
• Storage
• Benchmarks & evaluation
• Contributing
• Cite this list

How to use this list

Each project is tagged across three dimensions:

• Memory Tool — purpose-built to give agents persistent, retrievable memory.
• LLM Framework — general agent/app frameworks that include a memory module.
• Optimizer — improves retrieval, prompts, or embeddings rather than storing memory directly.
• Storage — the underlying database (vector store or graph DB) a memory layer is built on.

Types of AI memory

Most memory systems combine a few well-established kinds of memory:

• Short-term (working) memory — transient context held within the model's context window for a single task or conversation.
• Long-term memory — information persisted externally and retrieved across sessions. It's commonly broken down into:
  • Episodic — specific past events and interactions ("what happened").
  • Semantic — facts, preferences, and entities ("what is true").
  • Procedural — reusable workflows and skills ("how to do something"). Implementations differ mainly in how they store and retrieve this memory: vector search for semantic similarity, a graph for entities and relationships, or a hybrid of both (often with keyword/BM25 signals on top).

Memory tools

LLM frameworks

Optimizers

Storage

Benchmarks & evaluation

Common benchmarks used to compare AI memory systems on long-term recall and reasoning:

Memory tools often report scores on these (and on additional benchmarks such as ConvoMem and BEAM) — see each project's own README for current numbers.

Contributing

There might be many more companies and projects we aren't aware of. Your feedback and contributions are appreciated! ❤️

To add a project, please open a pull request including its name, URL, source model (Open source / Managed / Closed), GitHub URL (if any), category, and storage model (Graph / Vector / Graph, Vector).

Cite this list

If this list is useful for your work, please cite it:

@misc{awesome-ai-memory,
  title        = {Awesome AI Memory},
  author       = {topoteretes},
  year         = {2026},
  howpublished = {\url{https://github.com/topoteretes/awesome-ai-memory}}
}