> ## Documentation Index
> Fetch the complete documentation index at: https://docs.openp.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# ChatBox

> Connect OpenPAI in the ChatBox desktop client

[ChatBox](https://chatboxai.app/) is an open-source, cross-platform AI desktop client for Windows / macOS / Linux / iOS / Android.

## Configuration steps

<Steps>
  <Step title="Open settings">
    Gear icon at the bottom-left → **Settings**.
  </Step>

  <Step title="Choose a model provider">
    Under **Model Provider**, choose **OpenAI API** (or a custom one).
  </Step>

  <Step title="Fill in connection details">
    * **API Host**: `https://openp.ai/v1`
    * **API Key**: `sk-XXXXXXXXXXXXXXXX`
    * **Model**: pick from the dropdown or enter a custom model ID, like `gpt-5.5`, `claude-opus-4-8`
  </Step>

  <Step title="Save and start chatting">
    Back on the main screen, start a new conversation and pick the model at the top.
  </Step>
</Steps>

## Add multiple models

ChatBox lets you preset multiple models in the **model list**, so you can switch directly when changing conversations:

1. Settings → Models → Add custom model.
2. Enter the model ID from the OpenPAI console.
3. After saving, it appears in the dropdown at the top of the main chat screen.

A recommended model combination to preset:

* `gpt-5.4-mini` (everyday chat, cheap)
* `claude-opus-4-8` (code and long text)
* `gemini-3.1-pro-preview` (very long context)
* `o3-mini-high` (complex reasoning)
* `dall-e-3` / `gpt-image-1` (image generation)

## Mobile

The ChatBox iOS / Android apps also support a custom API Host;
the configuration flow matches the desktop. We recommend setting the API key to a sub-token with a quota cap, to limit losses if the device is lost.

## Notes

* ChatBox uses the OpenAI protocol by default, so the Base URL must include `/v1`.
* If ChatBox shows "cannot connect" / "Network Error", check:
  * Whether the network can reach `https://openp.ai`.
  * Whether there's an extra trailing slash on the Base URL.
  * Whether a system proxy is interfering with requests.
