Loading the catalogue…
The AI gateway for European companies
Every model, one API, one invoice
Connect your agents, apps and servers to Stav once. You get 100+ models from 23 Inference Providers at the providers’ own prices. The Smart Router picks a model that fits each request, so simple work doesn’t run on your most expensive model, and it switches to another provider if one fails. Every response shows which company ran it and in which country.
100+ models23 Inference ProvidersOne contract, one invoiceProviders’ own prices, plus a platform fee
from openai import OpenAI client = OpenAI( base_url="https://api.stav.ai/v1", api_key="sk_stav_live_…", ) client.chat.completions.create( model="auto", # Stav picks messages=[{"role": "user", "content": "Hello"}], )
In your request logMistral Small 3.2 · Scaleway · France
EU only · no retention · no training · €0.000008
EU only · no retention · no training · €0.000008
One place for every model
Stop managing a dozen AI vendors
Sign one contract and get one bill. Every model from every provider sits behind one API.
One contract, one invoice
Pay with prepaid credits from one wallet. See what you spend per model, router and API key. You don’t need an account with each AI company.
One key for every model
Try a new model by changing its name. No new sign-up, no new key, no new procurement round.
Works with what you already run
Stav accepts Chat Completions, Anthropic Messages and OpenAI Responses requests. Claude Code and Codex connect with one setting.
If a provider fails, the request moves to another
An error or a rate limit sends it to the next provider on the list. You don’t write the retry code.
Spend less on tokens
Don’t pay frontier prices for simple tasks
Most requests don’t need your most expensive model. Stav helps you send each one to a model that is good enough, and shows you what it cost.
Name a model, or let Stav pick
Send a model name and you get that model. Send auto and the Smart Router picks one that fits the request and your budget.
Compare before you commit
Every model page lists each provider’s price per token. Filter the catalogue by price and pick the cheapest one that does the job.
How much can be spent
Daily and monthly limits per workspace, key and router. A soft limit warns you; a hard limit stops requests.
Sorting 10,000 support tickets
One month. 600 tokens in and 150 out per ticket.
Catalogue prices on 15 October 2026, cheapest serving provider for each model. For simple work like triage, the small model is good enough. Hard tasks still need the bigger one.
Pricing
Pay only for what you use
Model prices are the providers’ own catalogue prices, open for anyone to read on each model’s page. A small platform fee is added when you buy credits.
No subscription
Buy credits and use them when you need them.
No per-seat fee
Add as many people and apps as you like.
Providers’ own prices
Read every price on the model’s page before you send a request.
From the team behind Jottacloud
“We built Jottacloud as a privacy-first alternative to the tech giants. Stav is that conviction applied to AI, where the stakes are higher. A model has to read what you send it, so encryption can’t protect your content. What protects you is control: you choose which providers and countries may process your data. Stav is an independent gateway. We don’t make models, we don’t train on your data, and we keep you from being locked in to one vendor.”
RRRoland RabbenFounder & Chairman, Stav AI
Send your first request
Create a workspace, get an API key and connect your agents, apps or servers to api.stav.ai. Add a card, and we add free credits to your wallet.