Hey hunt folks,
Built this because I am tired of re-explaining my code architecture to agents every 5 minutes. If you bounce between cursor and claude all day, you know that frustration where the agent "forgets" your db schema or the core logic you just spent two hours on.
Nucleus is an "agentic os" that lives right in your git repo. It turns your .brain/ folder into a shared identity...
It s basically:
We built Nucleus because we were terrified of our own agents. π‘οΈ
Last month, I gave an autonomous agent access to my filesystem to 'refactor some code'. It optimized my 'docker-compose.yml' by deleting it. That was the wake-up call.
We are entering the Agentic Age, but our infrastructure is still built for simple chats. Nucleus is my answer: an Agentic OS that runs 100% locally on your machine.
The Sovereign Pillars:
π‘οΈ Hypervisor: Intention-aware security that stops agents before they do damage.
π§ Engrams: Git-native long-term memory that lives in your project repo.
π Recursive Mounting: Orchestrate complex tool meshes (Stripe, Postgres, etc.) instantly.
No SaaS. No cloud. Just Sovereign AI. Iβd love to hear how youβre handling agent safety in your local loops!
π Deep Dives & Community:
πΊ Watch the Demo: https://youtu.be/D1B6m_F-h80
ποΈ Architecture: https://dev.to/nucleusos/how-i-synced-cursor-claude-and-windsurf-with-one-shared-brain-mcp-1mh4
πΎ Join the Tribe: https://discord.gg/RJuBNNJ5MT
π Launch Day Special: We are doing 1-on-1 'Agent Architecture Audits' for the first 10 people who leave a technical comment today. Let's harden your local workflows together!
β The Nucleus Developer
Thanks for checking Nucleus v1.0.7. We focused on three things:
Identity: Git-native engrams so context persists across sessions.
Security: Command approval + audit log; no hidden execution.
Setup:Β
Β now auto-detects Claude Desktop and sets config.
If you want help wiring Nucleus into your stack, drop your setup below. We will help the first 10 comments get perfectly configured.