Skip to content
souveraen.ai
DEAuf Deutsch ansehen
Developer API

Your knowledge and your agents, through an API.

An OpenAI-compatible interface on your own tenant. Your applications query models, receive sourced answers from your knowledge and start published agents. Access rights and approvals apply exactly as they do in the interface.

Start for free
See the endpoints
  • OpenAI-compatible
  • Hosted in Germany
  • Keys per person or workspace

chat.py

from openai import OpenAI

client = OpenAI(
    base_url="https://your-tenant.cust.souveraen.ai/v1",
    api_key="sk-svr-…",
)

answer = client.chat.completions.create(
    model="souveraen-auto",
    messages=[{"role": "user", "content": "Summarise the contract in three sentences."}],
)

The official OpenAI SDK with a different base URL.

Endpoints

One interface for models, knowledge and agents.

Tools and libraries that already speak the OpenAI interface only need a new base URL and a key.

Chat completions
With or without streaming, with tool calls and structured output. Pick the model yourself or leave the choice to souveraen-auto.
Answers from your knowledge
The model souveraen-wissen searches your connected sources and returns the citations as a field in the response.
Embeddings
Vectors for your own search and analysis, with the same key and the same metering.
Agents API
List, start, follow and cancel published agents. Only agents you switch on are reachable.
Signed callbacks
A run can report its result to your address, signed so you can verify it. The run is also available as an event stream.

Keys and control

Every key has an owner, a budget and a log.

A key can never do more than the person or workspace it belongs to. What it may do is set when you create it.

Security in detail
Personal keys
Act as the person, with their current access rights. When the rights change, so does what the key can see.
Workspace keys
A service identity for exactly one workspace, created by its managers. Suited to applications without a person behind them.
Scope and rate limit
Per key you choose whether it may use models, knowledge, embeddings or agents, and how many requests per minute are allowed.
Monthly budget
A key can have its own monthly budget of AI credits. When it is spent, the API answers with a clear error.
IP allowlist and expiry
Restrict a key to your address ranges and give it an expiry date. The secret is shown once.
Usage and audit log
Every call is counted and logged per key: endpoint, model and consumption, never content.

In every plan

The API is included, on the free plan too.

Calls consume the same AI credits as work in the interface. There is no separate API fee.

Per plan

OpenAI-compatible API
Agents API
Active keys per person
Requests per minute
Budget and limit per key
Usage visible per key

Free

Pro

Max

Yes
Yes
Yes
Yes
Yes
Yes
1
10
25 to 50
10
60
120 to 300
Yes
Yes
Yes
Yes
Yes
Yes
Start for free

Max: 25 keys and 120 requests per minute on Max 5×, 50 keys and 300 requests on Max 20×. Applies to On Demand.

Questions

What developers ask first.

On compatibility, rights and metering.

A specific project?

Tell us which system you want to connect and we will work out the right route. Phone +49 3744 365 2202.

Book a call

Models, chat completions with streaming and tool calls, and embeddings follow the OpenAI format, error messages included. You use the official SDK with a different base URL. The Responses interface is not offered.

Try it yourself

The first call needs one key.

Create an account, generate a key and put the base URL into your application. The API is included on the free plan.

Start for freeBook a call
  1. Create an account1
  2. Generate a key2
  3. Set the base URL3