01
Orchestration
The Hub assigns each job to a local or cloud model based on your rules — sensitivity, size, and task type. One decision layer instead of ad-hoc tool picking.
Product
Mainsches Hub combines model orchestration, file handling, workflows, and permissions in one workspace. This page explains the pieces and how they fit.
Product
01
The Hub assigns each job to a local or cloud model based on your rules — sensitivity, size, and task type. One decision layer instead of ad-hoc tool picking.
02
Local files are first-class citizens: the Hub can process documents on your own hardware, keep references to sources, and return results you can check.
03
Recurring work becomes a saved workflow with visible steps. Workflows are created from real tasks and refined over time — adaptivity is the point.
04
Every job carries explicit boundaries: which data, which actions, which model class. The system enforces them; no job silently widens its own scope.
Product
Convenience matters, but never at the cost of knowing what runs where. Defaults are conservative; expansions are explicit.
Local AI is not a demo mode. Core tasks are designed to work on your hardware, without a cloud account being a precondition.
A result you cannot reconstruct is a liability. Steps, scopes, and model choices are recorded so output can be defended.
What is planned is labeled as planned. Mail and service connectors are in development and communicated exactly that way.
FAQ
No. A chat can be part of the surface, but the product is the control layer underneath: model orchestration, workflows, permissions, and traceable execution.
The architecture is model-agnostic: local models on your hardware and cloud models within approved scopes. Concrete model line-ups depend on your setup and are discussed in the pilot program.
Primarily small and mid-sized companies that want AI leverage without losing control over their data — plus partners and investors who want to build that future with us.