Introduction

Sovereign is one endpoint for your company's AI traffic. Point your tools and apps at it, and each request goes to the model that does the work as well for less.

Your people, coding agents and apps keep sending the requests they send today, in Anthropic's or OpenAI's format. Sovereign answers them on the right model and shows you what every request cost, what it saved and what it sent.

How it works

1. Your client sends a request to https://sov.solibra.ai with a Sovereign key, in the format it already uses: Anthropic Messages, OpenAI Chat Completions or OpenAI Responses.
2. Sovereign chooses the model. The model your request names is the starting point. Sovereign moves to a cheaper model when it will do the work as well, and to a stronger one when the work needs it.
3. The answer comes back in the same format, streamed if you asked for a stream, with an x-sov-model header naming the model that answered.

Every request is recorded for your dashboard: who sent it, what it was for, the model that answered, what it cost and what it saved against the model it asked for. Sensitive data in requests is detected and shown, masked.

What you need

Your company's provider key, connected once on the Keys page: OpenRouter for every model, a provider's own key for its models, or your Amazon Bedrock, Azure AI Foundry or Google Vertex AI account. Then a Sovereign key for each person and app. That is the only key your people and apps ever use.

Base URLs