PRIVATE AI
Your AI.
Your infrastructure.
Most AI today is an all-or-nothing deal: everything you type goes to someone else’s cloud. DRENTIR is designed around a different idea — you choose where your intelligence runs and where your data goes, per task, per workspace, per user.
Design principles — illustrated, not production software.
YOUR CHOICE
Three ways to run your intelligence.
Not a setting buried in a menu — a first-class choice for every task.
On your hardware
Local-only
Capable models running on your own machines for work you want to keep close. Local tasks are designed to stay on your hardware — not the task, not the context, not the result leaves.
Suits: personal context, family data, financial documents — anything you prefer to keep local.
Routed per task
Hybrid
The default: each subtask goes where it belongs. Local for privacy and speed, cloud for capability — and DRENTIR shows you which is which before it acts.
Suits: everyday work where some tasks are sensitive and others simply need the best model.
Used deliberately
Cloud
When a task truly needs the most capable models, the cloud is used on purpose — not by default, not silently.
Suits: heavy reasoning, complex research, long documents — where capability outweighs locality.
For organizations
Private infrastructure
DRENTIR on infrastructure you control — your data center or private cloud, your keys, your retention rules. Designed for teams with compliance requirements.
Suits: companies that need data residency and auditability under their own roof.
DRENTIR is designed to minimize unnecessary provider-side data exposure and retention. Exact retention, deletion and training policies depend on the deployment mode and connected providers, and will be documented per integration.
DATA CONTROL
Data control by deployment mode.
| Aspect | Local-only | Hybrid | Cloud | Private deployment |
|---|---|---|---|---|
| Model execution | On your hardware | Per task: local or cloud | Cloud providers | Your infrastructure |
| Data residency | Your machines | You choose per task | Provider regions | Your infrastructure |
| Provider exposure | None by design | Only for cloud-routed tasks | Per provider policy | None by design |
| Credential control | You hold the keys | You hold the keys | You hold the keys; provider sees routed data | You hold the keys |
| Offline capability | Full | Partial — local tasks | None | Full |
| Admin control | You | You | Shared with provider | You |
DATA BOUNDARIES
You define what DRENTIR may see, remember, share.
Every piece of data the system touches is either inside your boundary or outside it — and you decide which is which.
Stays with you
- Local tasks and personal context
- Files and messages you mark private
- Voice conversations at home
Leaves only when you allow
- A task you explicitly route to the cloud
- A message you approve to send
- A lookup you ask the web to answer
01
Per user
Each person sets their own boundary. Your private context is yours — it does not flow into a colleague’s tasks, a family member’s requests, or any shared workspace.
02
Per workspace
Work, family, a side project — each workspace draws its own line. Data from one workspace is kept separate from another unless you explicitly carry it over.
The boundary is yours to move — per user, per workspace, per task.
PRINCIPLES
Privacy as architecture, not a promise.
No absolute claims — just design decisions that put control in your hands.
Local first
Work stays on your machines unless there is a reason for it to leave — and the reason is shown to you before anything moves.
No training on your data
DRENTIR is designed so your tasks, files and conversations are not used to train models. Exact training and retention policies depend on the deployment mode and connected providers, and are documented per integration.
You hold the keys
Encryption and access control answer to you, not to a vendor dashboard. Revoking access removes it from the places you control.
Visible data flow
The activity trail shows what DRENTIR touched and where it went. No black boxes between your instruction and the result.
FAQ
Questions about privacy.
Straight answers. Where something is not decided yet, we say so.
DRENTIR is designed so your tasks, files and conversations are not used to train models. Local tasks stay on your machines. For cloud-routed tasks, exact training and retention policies depend on the connected provider and are documented per integration.
Final hardware requirements are not set — DRENTIR is in development. The direction is clear: capable models that run on a modern home computer or a small home server for everyday tasks, with hybrid routing to the cloud when a task genuinely needs more. Early access will define the exact specs.
You always see the routing before it happens. In hybrid mode each subtask is labeled with where it runs and why. The activity trail records what left, where it went, and what came back. Nothing leaves silently.
Yes. You can review what DRENTIR remembers per workspace and delete it — selectively or entirely. Since you hold the keys, deletion revokes access everywhere you control. Exact deletion behavior for connected providers is documented per integration.
Questions about data boundaries?
Tell us what worries you — privacy is the part we like talking about most.