Swiss AI Control LayerTrustRouteCredits tracked

Trust & infrastructure

See what RouterLab hosts, and what proprietary providers process.

This page separates roles: RouterLab controls access, credits, and usage, and hosts open-source models in Switzerland and Germany. Proprietary providers process requests sent to them.

TrustBoundaryMap

Responsibility map

A request crosses several zones. RouterLab makes those zones readable.

RouterLab API · CH / DE

Client source

App / agent / workflow

OpenAI SDK
Claude Messages API
Product automation

ROUTERLAB CONTROL

RouterLab API

Authenticates API key
Applies permissions and access
Attaches usage and credits

Model processing

According to model type

Open source: RouterLab CH / DE
Proprietary: external provider
OpenAI / Anthropic / Google

Visible

Published route + hosting

Measure

Usage + credits

Metadata

No prompt content

What RouterLab controls

Access, routes, credits, and usage are grouped in one place.

RouterLab centralizes API access, applies account permissions, shows the published logical route, and ties each request to credits. Internal routing logic remains confidential.

Visible provider

Routing

Destination and format visible

Each published route shows the call format and whether the model is hosted by RouterLab or processed by a third-party provider, without exposing internal routing logic.

Credits tracked

Credits

Consumption tied to balance

Each request feeds usage and the credit ledger so real cost can be tracked.

RouterLab key

API keys

One key in the application

Your apps call RouterLab with a RouterLab key instead of storing provider keys everywhere.

Dashboard

Tracking

Readable usage after the call

The dashboard shows the published logical route, consumption, and remaining credits, without displaying prompt and response content or internal routing logic.

Infrastructure and regions

The CH / DE signal is visible, and third-party providers are too.

RouterLab infrastructure and open-source models are hosted in Switzerland and Germany. Proprietary models are processed by their respective providers.

CH / DE infrastructure

RouterLab and open source in CH / DE

RouterLab infrastructure and open-source models run on our servers in Switzerland and Germany.

External processing

Named proprietary providers

OpenAI, Anthropic (Claude), and Google (Gemini) process requests that use their models under their own infrastructure and policies.

Visible routes

Readable provider statuses

Published routes expose format, status, and hosting category. They do not describe internal routing logic.

TrustRoute

Explicit hosting and providers

ProviderFormatRoleWhat it meansSignal
OpenAI
OpenAI SDKThird-party providerProcesses the request on OpenAI infrastructure under its own policiesThird-party provider
Anthropic (Claude)
Claude Messages APIThird-party providerProcesses Claude requests on infrastructure external to RouterLabThird-party provider
Google (Gemini)
OpenAI-compatibleThird-party providerProcesses Gemini requests on Google infrastructureThird-party provider
Open source (RouterLab)
OpenAI-compatibleHosted by RouterLabModels run on RouterLab infrastructure in Switzerland and GermanyHosted by RouterLab

Data, keys, and limits

Trust also comes from what is limited.

The right trust level is precise: state what is controlled, and do not promise what depends on a provider.

Centralized access

API keys

Your integrations call RouterLab with a RouterLab key. Provider keys are not distributed into your client apps.

Credit ledger

Usage and credits

RouterLab keeps only the events and metadata needed to track credits, costs, usage, and support.

No durable archive

Prompts and responses

RouterLab does not log or durably archive their content. A very short-lived technical cache may be used when strictly necessary to operate the service.

Roles

Account access

Customer and admin areas remain separated by authentication and application roles.

TransparencyBoundary

Explicit limits

This page avoids unverified legal promises. It describes RouterLab's operational role.

Not a provider retention promise

OpenAI, Anthropic, and Google processing and retention policies depend on the provider and selected proprietary route.

Not a legal certificate

This page explains operational behavior. Legal commitments must remain in legal documents or contracts.

No provider hiding

When a route goes to a third-party provider, RouterLab should make it explicit instead of burying it.

Legal pages remain the source for terms, privacy, and data protection.

DataFlowDiagram

Trust follows the same path as the request.

Tracking covers the operational metadata needed for permissions, the published destination, usage, and credits. It includes neither prompt and response content nor internal routing logic.

01

Request

POST /v1/chat/completions

02

Control

Key + permissions + plan

03

Destination

Hosting / provider

04

Credits

Ledger + usage

Verification

Verify the flow with a trial key.

Test the endpoint, choose a model, and watch credits and usage return in the dashboard.