Partner integration · Live
Infercom Router
EU-sovereign AI models directly in Claude, ChatGPT and the command line — one connector, one login, cost per call visible.
Anyone who wanted to use European models previously had to build their own application. The Router makes them usable in the assistant that teams already use — and along the way decides which model can handle the task and what it costs.
The problem
Three operational problems at once: how do you get EU models into the existing assistant? Which model can handle images, tools or long contexts, and what does it cost? And how do you keep pace with a model catalogue from which models keep disappearing?
How it works
- 01 One connector, one login: enter the MCP address in the assistant, sign in once (OAuth against our own identity management), done.
- 02 Routing rules instead of guesswork: the default is EU-sovereign. Images, audio, very long inputs and tool calls go automatically to the model that can handle them; a cost limit per request picks the cheaper one where needed.
- 03 Own key or credit balance: anyone with their own Infercom access stores it encrypted and is billed directly there. Everyone else pays from a credit balance.
- 04 Daily catalogue reconciliation: the model catalogue is compared against the previous day every day; the connector does not go stale.
- 05 Measurement per call: model, tokens, cost and response time are recorded and shown in the usage report. Inputs and answers are not stored.
Sovereignty & evidence
- Operated in Germany on STACKIT with own database, own key store and own identity management — no third-party login provider.
- Inference via Infercom: EU models in German data centres, Zero Data Retention. Non-European models are never selected automatically.
- Only what is needed for attribution and billing is stored: user mapping, encrypted keys, usage events. No content.
- Access with revocable tokens (introspection per RFC 7662); fully handoverable as code.
Why an integrator builds a connector
Integration means making things available where people work. Most teams today work in Claude or ChatGPT. The Router is the bridge from there into European data centres — and the pattern by which we also plug corporate knowledge, specialist systems and verified sources into assistants.
Partnership
Inference comes from Infercom, a provider headquartered in Luxembourg with data centres in Germany. Pioneerdesk acts as Solution Partner. The Router is built so that it can be operated and handed over as a standalone package.
Status
In productive use, and at the same time a young project: the model catalogue changes often, so we only quote figures about it with a date. Public access runs via suveri.de.
Frequently asked questions
- What does the Router have to do with Suveri?
- It is one codebase with two faces. The Router is the connector and delegation layer (model selection, key custody, measurement, billing). Suveri is the end-user workspace on top of it — and can itself be used as a connector in turn. Pioneerdesk operates both as Solution Partner of Infercom.
- Why not use the Infercom API directly?
- You can. The Router saves you three things: building your own application, choosing the model per task, and constantly maintaining a catalogue that keeps changing.
- Who sees my inputs?
- Infercom processes them with Zero Data Retention; the Router does not store them. What remains is user mapping, encrypted keys and usage metrics — nothing else.
A comparable challenge on your desk?
Send us three sentences. We reply within one business day with an honest assessment — including when we are not the right partner.