Documentation

Everything
an integration needs.

From the first call to running your own agents. The API is OpenAI-compatible, the models run on machines inside the European Union, and every machine-readable description is published at the gateway api.simosphereai.com without a login.

Quick start in two lines.

If you already build against an OpenAI-compatible API, change two things: the base URL and the key. Your existing integration keeps working.

Every call goes through a single base URL. It serves chat completions along with all other operations; there is no second endpoint for special cases.

Replace the key in the example with your own. It belongs in an environment variable, never in source code. Not even temporarily.

First call with curl
curl https://api.simosphereai.com/v1/chat/completions \
  -H "Authorization: Bearer sk_live_XXXXXXXX" \
  -H "Content-Type: application/json" \
  -d '{"model":"mistral-small-latest","messages":[{"role":"user","content":"Hello"}]}'
Base URL
https://api.simosphereai.com/v1

Where the material lives.

Four entry points, depending on what you are after.

  • Specifications and manifests

    OpenAPI 3.1, MCP manifest, server card, OAuth discovery, interface catalog. Six addresses, reachable without a login, each one verified.

  • The MCP server

    What the server exposes, how it runs on your own premises, and which registries carry its listing.

  • The connectors

    Which systems can be linked up: enterprise software, financial services, manufacturing and media.

  • Compliance and data processing

    What to settle before going live: the data processing agreement, certification status, EU AI Act.

The platform at a glance.

Four properties that hold for every integration.

  • One point of entry

    The gateway at api.simosphereai.com handles authentication, per-key limits, billing and routing across the model backends. Outwards it presents an OpenAI-compatible API that any existing client library can talk to.

  • Data sovereignty

    The models run on machines in Germany. Request data does not leave the European Union. Prompts and completions are neither stored nor used for training.

  • Model choice

    Open-weight models behind a single API. Switching costs one parameter, not a code change. Each model runs in its own isolated environment.

  • Running it at scale

    Per-tenant key management, per-key limits, usage reporting, notifications and role-based access control, all with complete audit logging throughout.

Questions this page does not answer?

For anything beyond getting started, a conversation is the shorter route than another page.