Local processing where it fits
Documents and files can be processed on your own hardware. For sensitive material, the data simply does not leave.
Security
Security and privacy, designed in from the start.
Most AI tools ask for trust. Mainsches Hub is built to replace trust with structure: local processing where it fits, explicit boundaries, and traceable execution.
Security
The security model
Explicit job boundaries
A job is a contract: defined input, defined actions, defined model class. Nothing widens itself.
Permissions & scope
Scopes are enforced by the system. An approved cloud step stays exactly that — one step, one scope.
Traceable workflow steps
Every step is recorded. You can reconstruct what ran, where, and with which boundaries.
Privacy by design
Data minimization is structural: jobs receive the context they need, not your whole environment.
Cloud without coercion
Cloud capacity is available when you choose it. It is an option with a scope — never a hidden default.
Security
How data is treated
Three plain rules describe the intended data path in Mainsches Hub.
- 01
Data stays local by default
The default path for files is your own hardware. Leaving it is an explicit, scoped decision.
- 02
Cloud steps are scoped
When a job uses cloud AI, the scope defines what is shared — and that scope is visible to you.
- 03
Results are accountable
Outputs carry their history: steps, model class, and boundaries. You can always answer “where did this come from?”
This page describes the product's architecture and design goals. Mainsches does not claim external certifications, audits, or compliance attestations at this stage.
FAQ
Security FAQ
Does my data have to go to the cloud?
No. Local processing is a core capability, not a fallback. Cloud models are used only when you approve them for a scoped step.
Can I see what a workflow did?
Yes. Steps are recorded with their boundaries and model choice, so execution can be reconstructed after the fact.
Is Mainsches Hub certified?
Mainsches Hub is in development and does not claim external certifications at this stage. The security model is architectural: local processing, scopes, and traceability.