Model routing
One API for the whole company. Behind it, many models from different vendors. You route by task and see costs as you go — with limits that stop invoice surprises.
What the model gateway is
A central point between your apps and the LLM world. Nextcloud, internal scripts, and the assistant connect here.
The API format works with common SDKs. You do not rewrite code when you change models. Usage tracking and spending limits are part of the gateway, not an add-on.
Many models at once
OpenAI, Anthropic, open source on your server, European providers. Switch by job: a fast model for classification, a stronger one for contracts.
One address
Apps call one endpoint. Vendor keys live on the server, not in every developer’s laptop.
Fallback routes
When the primary model is down or overloaded, the request goes to a backup model. Fewer outages in production.
Cost tracking
Every request is counted: user, team, model, project. See what chat, contract summaries, or automation cost — not three vendor invoices later.
Spending limits
Monthly cap per department or whole company. Alerts before you hit the limit, hard stop at a critical threshold. One team cannot burn the budget over a weekend.
Policy and EU
We set which models each team may use. The gateway runs in the EU. Logs under your contract.
Typical uses
- Nextcloud AI apps via Nextcloud integration.
- Internal tools (helpdesk, CRM, scripts) without calling US APIs directly.
- Trying a new model on part of traffic before full rollout.
- Budget control: per-team limits, alerts, and one invoice from us.
Billing and audit details are on costs and limits. Employee chat is on private assistant.