28naem-del

28naem-del / mnemosyne

Public

Cognitive Memory OS for AI Agents — persistent, self-improving, multi-agent memory

mnemosy.ai agents ai ai-agents artificial-intelligence cognitive
18
3
100% credibility
Found Feb 27, 2026 at 11 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

Mnemosyne is a brain-inspired memory system for AI agents that enables persistent storage, intelligent recall, self-improvement through feedback, and collaborative knowledge sharing across multiple agents.

How It Works

1
🔍 Discover Mnemosyne

You hear about Mnemosyne, a smart memory helper that lets your AI friends remember conversations, learn from chats, and share knowledge like a real brain.

2
🚀 Set up your AI's memory

With a quick and easy start, you connect it to your AI so it can begin storing and recalling important details from your talks.

3
💾 Save your first memory

You share a fact or preference with your AI, and Mnemosyne smartly files it away, linking it to related ideas for easy future access.

4
🧠 Ask and get smart answers

When you chat with your AI, it pulls up the right memories automatically, giving helpful context that feels personal and aware.

5
👍 Help it learn better

You give simple thumbs up or down on the memories it recalls, and it gets smarter over time, remembering what works best for you.

6
🤝 Team up multiple AIs

Your AIs start sharing memories across the group, spotting patterns together and building collective smarts without starting over.

✨ Your AI feels alive

Now your AI companion remembers you, learns from every chat, avoids old mistakes, and works seamlessly with others—like having a thoughtful friend.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 11 to 18 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is mnemosyne?

Mnemosyne is a TypeScript library that builds persistent, brain-like memory for AI agents, handling storage, retrieval, and self-improvement without LLM costs during ingestion. Agents store facts, preferences, or procedures via a simple API—`m.store("User likes dark mode")`—then recall with multi-signal ranking including decay, confidence, and graph links. It tackles agent amnesia using Qdrant for vectors, Redis for caching/pub-sub, and FalkorDB for temporal graphs, enabling cognitive AI memory on GitHub that evolves over time.

Why is it gaining traction?

Unlike Mem0 or Zep, which rely on pricey LLM calls per memory (~$0.01 each), Mnemosyne runs a zero-LLM 12-step pipeline in under 50ms, with built-in features like activation decay, reinforcement feedback, and cross-agent synthesis. Developers love the free knowledge graph, procedural memory immune to forgetting, and tools for proactive recall or theory-of-mind queries across agent meshes. It's a full cognitive architecture on GitHub that scales to production without vendor lock-in.

Who should use this?

AI engineers building multi-agent systems for devops, customer support, or research assistants, where agents need shared expertise like "DevOps agent knows DB configs." Solo agent devs facing cognitive memory issues in long sessions, or teams prototyping cognitive tools on GitHub wanting to avoid cognitive overload from stateless chats. Ideal for cognitive memory exercises in agent training without the cognitive memory impairment of basic vector stores.

Verdict

Try it for agent prototypes—quick Docker setup yields sub-200ms recall on 13k memories—but temper expectations with 10 stars and 1.0% credibility signaling early maturity. Solid docs and tests make it worth forking for custom cognitive services, despite lacking polish for enterprise yet.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.