01
Memory lock-in
The more useful an AI becomes, the harder it is to leave, because your history, preferences, projects, and context stay trapped inside one platform.
Sovereign AI for the People and communities
Private memory for the age of AI.
Mana gives you private, portable memory — so your intelligence can be powered by any local or remote model without surrendering your history to one platform.
The Problems
Convenience today can mean loss of control tomorrow. Centralized AI platforms create risks you cannot afford to ignore.
02
Privacy dependency
When your continuity depends on one platform, you may be pushed to accept terms, tracking, data retention, or privacy tradeoffs just to keep access to your own AI history.
03
Identity capture
AI memory can become a detailed map of your life, including your work, relationships, values, fears, desires, habits, and vulnerabilities.
04
Data Exploitation
Your AI history can be harvested to profile, predict, persuade, and influence you more precisely.
Mana’s counter-model
Portable Memory, Selectable Intelligence
Mana separates your memory from the AI model. Your memory becomes portable continuity, not a feature trapped inside one provider.
User-owned memory
Your context belongs to you, with control over what is kept, used, exported, or withheld.
Selectable intelligence
Use the model that fits the task without surrendering continuity to a single provider.
Portable context
Carry projects, preferences, decisions, and working history across tools instead of restarting.
Limited disclosure
Share only the context a task needs, instead of exposing your full memory trail by default.
Long-term continuity
Your AI life can develop over years, while models remain interchangeable tools around your memory.
Continuity across models
Your memory stays yours. The AI is your choice.
Mana preserves the context that makes AI useful, then lets you route it to the model you choose.
Your memory
Chats, notes, files, projects, and ideas.
AI models
Compare answers. Switch anytime.
Choose how close the intelligence stays to you.
Three layers of AI sovereignty
Mana begins by separating your memory from the model. Your context stays portable, while intelligence becomes selectable.
Some tasks need the strongest cloud models. Some need the privacy of a local model. Some communities may need their own private nodes. Mana is designed to support all three paths.
01
Selectable Intelligence
Use ChatGPT, Claude, Gemini, Grok, open models, or other providers without locking your memory inside any one platform.
02
Local LLM
Route sensitive work to models running on your own device through tools like Ollama or LM Studio when privacy matters more than maximum cloud power.
03
Private Nodes
Connect to user-owned, team-owned, or community-owned compute nodes for more power, shared infrastructure, and reduced dependence on centralized AI companies.
More sovereignty today, stronger sovereignty over time.
Mana does not claim that cloud AI is completely private. It changes the default pattern by limiting how much of your history any single model receives.
What Mana can improve now
Less total-history exposure
When you use hosted AI, the model may still see your current prompt and the selected context sent to it. But it should not automatically receive your whole history, full memory, or long-term identity map.
Why it matters
A harder-to-complete profile
By separating memory from the model and sending only limited context per request, Mana makes it harder for any single LLM provider to build a complete picture of your life, history, and identity.
Your memory belongs to you. The model provides temporary intelligence.
The honest promise is not that no data ever leaves. The promise is that your memory is not surrendered as the permanent property of one AI platform.
Context gate
Mana is your privacy guardian.
Most people using AI are forced to manually sanitize everything they paste - removing names, stripping company details, hiding personal or sensitive information, or deciding whether it's even safe to send at all.
Mana's Context Gate changes this. Before any context leaves your device, it inspects what you're about to send and automatically decides:
- What the model actually needs to know
- What should be summarized instead of sent raw
- What should be held back entirely
- What stays completely private on your device
You no longer have to manually protect your privacy every time you use AI. The Context Gate acts as an intelligent boundary - letting you get useful assistance while controlling exactly how much of your life, work, or data reaches the model.
Private memory
User rules
Model receives
Only needed context
Early access
Carry your memory forward.
Join the early circle shaping Mana.