Ready to get started?
Deploy sovereign AI on your infrastructure - in weeks, not months.
Katonic Token Gateway · 3:02
Group IT applies the licence, registers the provider accounts, entitles models per department and sets the cap; a department runs on it; a developer makes the call; the same cluster grows into the Token Factory.
What is said, chapter by chapter
The Katonic Token Gateway in three minutes: one governed gateway over the model accounts you already own.
Licence
It starts with the licence key. Group IT applies it on the console's Licence page: the gateway, the hosted provider accounts, guardrails and audit, with a free token allowance signed in. Nothing to serve, nothing to buy first: it starts on the accounts the company already holds.
Console
Group IT's console. The licence decides the menu: departments, charge-back, Provider Accounts, LLM Models and audit. Tokens managed is the meter: every provider, every department, against the free allowance. Provider Accounts is the door: Azure OpenAI, Bedrock and OpenAI, on Group IT's own write-only keys. Each department is a cost centre with a monthly cap. Legal has hit its cap and is refusing calls. On Entitlement and cap, Group IT picks the models a department may call. Anything else is refused. The cap is in dollars at list price: notices at 75 and 90 percent, a refusal at 100. LLM Models holds each entitled model's rate. A model with no rate is refused until Group IT types one.
Daniel runs Customer Service. Sam builds on it.
Now the department that spends the tokens.
Seat
Daniel runs Customer Service. His Control Room dashboard shows spend against the cap, requests, tokens and refusals. Every API key has a named owner, its tokens and its charge-back. Budgets shows the cap Group IT set: 78 percent used. Group IT's Halden baseline guardrail profile is bound to every department. The Audit Log shows every change to the department's keys, members, cap and entitlement. Studio on this licence shows Develop only, nothing greyed out: the licence decides what you see. Sam, a developer, lands on his Developer Hub keys: one personal key, with an expiry.
The call
The call is code: SDK & snippets gives Sam the client and the base URL. Group IT's administrator proves the rails: the guardrail Test runs a prompt injection carrying a national ID. Blocked in seventeen milliseconds: the injection is stopped and the ID redacted, before any model sees it.
A new licence key on the same cluster adds serving.
And when your own GPUs arrive?
Grow
A new licence key on the same cluster adds serving: the seat gets its Deploy section, LLM Deployment and Model Registry. The accounts, departments, keys and audit stay. The licence key is the switch.
That is the Katonic Token Gateway: one door, one meter, guardrails on every call, and room to grow. Start at katonic.ai.
Chapters
Recorded 2026-10-11 from a real build; the screens show a fictional company.
