Build with sovereign AI inference
One API for the OpenAI and Anthropic SDKs. Send model="auto", and every answer tells you the model, the Inference Provider and the country that served it.
- Make an API keyIn Your workspace → Connect → API keys.
- Point your SDK at StavChange the base URL. Your code stays the same.
- Read the receiptEach response carries
X-Stav-*headers: model, country, cost.
Python
from openai import OpenAI
import os
client = OpenAI(
base_url="https://api.stav.ai/v1",
api_key=os.environ["STAV_API_KEY"],
)
r = client.chat.completions.create(
model="auto",
messages=[{"role": "user", "content": "Hei, Stav!"}],
)
print(r.choices[0].message.content)
MCP server and Skills
Run Stav from Claude, Codex or Cursor
Paste one prompt into your agent. It connects the Stav MCP server to your workspace, installs the four Stav Skills, and checks that both work. Setup costs nothing.
- /move-to-stav
- Move an app, or one feature, onto Stav.
- /find-a-model
- Find the model that fits a task and your policy.
- /set-policy
- Choose where your requests may run.
- /cost-review
- See where your credits go, and spend less.
Works with
Claude
Codex
- Cursor
Hermes
Get started
Mint a key, point your SDK at api.stav.ai, ship your first completion in five minutes.
6 pages
Concepts
How Stav works: the sovereignty ladder, the Smart Router, reasoning, tools and structured output, credits and billing.
9 pages
Guides
Move from the OpenAI or Anthropic SDK, enforce a sovereignty floor, and run coding agents on Stav.
12 pages
API reference
Every endpoint, every field, every error, with code in curl, Python and Node.js.
33 pages