# Use Omnious with your LLM

Omnious lets you use your existing LLM application while looking for a lower-cost way to run its
requests. You do not need to change your prompts, learn a new model API, or choose an inference
provider yourself.

If your application already uses an OpenAI-compatible provider, setup means changing three values:

1. **Base URL:** the Omnious inference API URL.
2. **API key:** an Omnious inference key.
3. **Model:** an Omnious preset, such as `preset/coding`.

Ask your organization administrator for these values. Keep the API key private and store it in your
application's secret settings.

## Set up your LLM

Open your LLM application's provider settings and enter:

```text
Base URL:  https://<omnious-host>/v1
API key:   <your Omnious inference key>
Model:     preset/coding
```

The labels may be different in your application. Look for settings named **endpoint**, **API base
URL**, **token**, **secret key**, or **model**.

Then send a normal chat message. Omnious accepts the same OpenAI-compatible requests your
application already uses, including streaming, tools, and conversation history.

## Why this can cost less

Omnious makes your request available to participating providers. Providers return prices for the
work they can handle, and Omnious chooses a valid low-cost option for the preset you selected.
Continuing conversations may also reuse a provider's cache commitment, reducing the cost of repeated
context.

You do not need to compare providers or change your application when provider prices change. Keep
using the same Omnious base URL and preset.

> The current MVP is focused on routing. Billing and provider settlement are not yet part of the
> user-facing flow.

## If something goes wrong

* **Your preset is not available:** ask your organization administrator which preset to use.
* **No provider is available:** try again later or contact the person who manages your Omnious
  environment.
* **Your key is rejected:** create or request a new inference key.
* **A response stops while streaming:** the selected provider may have failed. Omnious does not
  silently switch providers halfway through a response.

When contacting support, include the `x-omnious-request-id` response header if your application shows
it. Do not share your API key or prompt.

## Keep your existing setup

Once configured, your application can continue to use its normal:

* prompts and conversation history;
* streaming or non-streaming responses;
* tools and structured output; and
* OpenAI-compatible client library.

You only need to change the Omnious settings when you switch environments or receive a new key or
preset.

If you provide inference capacity, [get started as an inference provider](/providers).