# Apps

https://omnimodel.me/docs/apps

Almost any app that speaks the OpenAI-compatible API can connect directly. The [app gallery](/apps) has step-by-step setups for popular chat clients, coding assistants, workflow tools, browser extensions and official SDKs, and fills in your key after you sign in.

## The three settings you need

Nearly every app asks for just these:

| Setting | Value |
| --- | --- |
| API host (Base URL) | `https://omnimodel.me/v1` |
| API key | A `key_...` key created on the console Keys page |
| Model | A model ID from the [catalog](/catalog) |

Choose **OpenAI-compatible** as the provider type (some apps call it OpenAI API Compatible or Custom OpenAI). For tools that only speak the Anthropic or Gemini protocol, use `https://omnimodel.me/anthropic` or `https://omnimodel.me/v1beta` instead.

## Choosing by app type

| Type | Typical use | Tip |
| --- | --- | --- |
| Chat client | Everyday chat, switching models | Give the client its own key and limit its models |
| Coding assistant | Code generation, refactoring, explanations | Pick a long-context model and set a daily spend limit |
| Workflow | Batch jobs, automation nodes | Set a monthly limit and watch the failure rate in [usage](/docs/usage?lang=en) |
| Browser extension | Page translation, quick explanations | Frequent calls, so prefer cheaper models |
| SDK | Calls from your own backend | Keep the key in server-side environment variables, never in frontend bundles |

## When the app cannot connect

1. Make sure the API host ends with `/v1`, without `/chat/completions`.
2. Make sure the key was copied in full and is not paused or revoked (error `auth.invalid_key`).
3. Check the model ID for typos, and that it is in the key's allowed models (`model.unavailable`, `auth.forbidden`).
4. If the app's model list is empty, check the key's allowed models: `GET /v1/models` returns only models that key can call.

## FAQ

### Should the API host include /chat/completions?

No. Use https://omnimodel.me/v1; the app adds the endpoint path itself.

### The app shows no models. What should I check?

GET /v1/models returns only models your key can call. Check the key's allowed models, or type the model ID manually.

### Can I use my key in a web page or extension?

In personal clients and extensions, yes, but never in publicly shipped frontend code. Give each app its own key with a spend limit.