ClaudioDrews

ClaudioDrews / memory-os

Public

A 6-layer memory operating system for Hermes Agent — persistent memory with Qdrant, structured facts, fabric recall, auto-curated wiki, and surgical context injection. Runs locally, any LLM provider.

22
0
80% credibility
Found Jun 01, 2026 at 22 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Memory OS is a local memory system for AI assistants that provides six layers of persistent memory, allowing the assistant to remember projects, decisions, and context across sessions without cloud subscriptions or vendor lock-in.

How It Works

1
💭 Your AI assistant keeps forgetting

You spend hours teaching your AI assistant about your projects, and the next day it acts like a stranger who has never met you.

2
🔧 You install Memory OS

With one setup, you connect your AI assistant to a personal memory system that runs entirely on your own computer.

3
🧠 Six layers of memory come online

Your assistant gains workspace files, conversation history, structured facts, cross-session memory, a searchable knowledge base, and an auto-organizing wiki.

4
✨ The first time it remembers

Weeks later, your assistant brings up a decision you made together, complete with context you never had to repeat.

5
📚 Your knowledge grows automatically

Every conversation, decision, and discovery gets captured, organized, and made searchable without you lifting a finger.

🤝 Your assistant becomes a true collaborator

It knows your projects, your preferences, and your history — like working with a colleague who was there for every conversation.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 22 to 22 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 memory-os?

Memory OS is a persistent memory layer for Hermes Agent that gives your AI assistant a real long-term memory. Instead of starting every conversation from scratch, it remembers your projects, decisions, and reasoning across sessions. The system uses six distinct memory layers working together: workspace files injected into every prompt, full-text search across conversation history, structured facts with trust scoring, cross-session recall via Fabric, a hybrid vector database combining dense embeddings with BM25 sparse search, and an auto-curated wiki that organizes knowledge continuously. Built in Python with Docker infrastructure (Qdrant, Redis, ARQ Worker), it runs entirely on your local machine and works with any LLM provider Hermes supports.

Why is it gaining traction?

The core insight here is that most memory solutions are cloud-locked or too shallow to matter. Memory OS solves both by running locally with no subscription and by implementing actual intelligence in how context gets injected. The surgical retrieval system uses relevance thresholds and per-session deduplication to avoid flooding the context window. The reflection engine actively reviews memories, detects contradictions, and updates confidence scores over time. For developers who want their agent to evolve rather than reset, this is the architecture that makes it possible.

Who should use this?

Developers who take Hermes Agent seriously and want an agent that actually remembers work across sessions. Teams running multi-agent setups where agents need to hand off context to each other. Anyone who has spent hours configuring an agent only to find it acts like a stranger in the next session. The training data extraction and model replacement pipeline also appeals to users who want to fine-tune a cheaper model on their own agent history.

Verdict

Memory OS is a serious, well-architected solution to a real problem, but with 22 stars it is early-stage software. The documentation is thorough and the six-layer design shows careful thought, but you should expect to do some integration work. The credibility score of 0.8% reflects this maturity level. If you are already running Hermes and frustrated by amnesiac agents, this is worth the setup effort. If you want something plug-and-play, wait for the project to mature.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.