Zanii LLM
Menu

INFERENCE YOU CAN PROVE

Every call. Provable.

The OpenAI-compatible API with proof built in. Run your models, keep your workflow, and get a signed receipt for every call.

OpenAI compatible / Billed in AED / Signed receipts

From an API request to a signed receipt on a public ledger POST /v1/chat/completions { "model": "…", "messages": […] } YOUR REQUEST ledger.zanii.agency PUBLIC LEDGER SIGNED RECEIPT
Illustrated request flowPrivate prompts. Verifiable proof.

What you get

One line to switch

Change the base URL in the OpenAI client you already use. Chat completions, streaming and tool calls behave the way your code expects.

A receipt per call

Signed, hash-chained, anchored in a public Merkle log. Hand the hash to an auditor and they verify it without trusting us.

The model is named in the signature

The receipt records which model served your call, so you can prove you were given what you paid for and not something cheaper.

Your prompt stays yours

Only a salted hash is published. The prompt and the answer never leave your systems, and a short prompt cannot be guessed from the log.

Billed by the token, in dirhams

Metered from the model's own token counts, charged in AED, with an invoice that can be re-derived from your usage at any time.

Hosted on our own hardware

We run the models ourselves. No reselling, no surprise upstream, and capacity we can answer for.

How it works

  1. Get a key

    Sign in with Zanii ID and issue an API key in the console.

  2. Change one URL

    Point your OpenAI client at llm.zanii.agency/v1.

  3. Call a model

    Chat completions, streaming, tool calls: the usual shapes.

  4. Get the receipt

    Each response carries its receipt hash and a verify link.

  5. Prove it later

    Anyone can check the proof on the public ledger, offline.

Pricing

AED / 1M TOKENS
ModelInput / 1MOutput / 1MContext
glm-4.7-flash0.251.50202,752
qwen3.8-27b0.754.50262,144

Prices exclude 5% UAE VAT, which applies to UAE customers.

Pricing →

Who it is for

Companies whose AI use has to stand up to a regulator, an auditor or a client: healthcare, banking and finance, insurance, legal, tax, government, logistics and anyone building on top of a model they have to answer for.

FAQ

Is it really OpenAI-compatible?

Yes. Point the OpenAI client at https://llm.zanii.agency/v1 and change nothing else. Chat completions, streaming and tool calls all behave the same way.

What is a receipt?

A signed record of one call (which key, which model, when) and hashes of the prompt and answer, written to a public Merkle log. Anyone you show it to can verify it offline, without trusting Zanii.

Do you see my prompts?

They pass through the gateway to reach the model and are not stored, and they are never published. The ledger only ever holds a salted hash.

How am I billed?

By the token, in AED, from the model server's own counts. Credit is prepaid, a call that cannot be paid for is refused before the model is asked, and the monthly invoice can be re-derived from your usage records.

Where do the models run?

On hardware we operate ourselves, not resold from another provider.