# Chatbox

> Connect Chatbox on every platform (Windows / macOS / Linux / iOS / Android / Web) to the LMU AI API — dual-protocol, a direct overseas connection.

URL: https://docs.lmuai.ai/docs/tools/chatbox







Chatbox is a multi-model AI chat client that covers every platform — Windows / macOS / Linux / iOS / Android / Web.

<Callout type="info" title="On mobile (iOS / Android)? Start here">
  Step-by-step setup for Chatbox mobile — including a configuration screenshot and the "Improve Network Compatibility" switch — is on its own page: &#x2A;*[Chatbox Mobile](./chatbox-mobile)**. The rest of this page applies to both desktop and mobile.
</Callout>

***

## Download and install [#download-and-install]

* **Official site**: [https://chatboxai.app](https://chatboxai.app)
* Download and install the version for your platform — no account needed to use BYOK (bring your own key) mode

***

## The two protocols [#the-two-protocols]

LMU AI supports two connection protocols; choose one based on the model type:

| Protocol                              | Models                                                          | API Host               | API Path        |
| ------------------------------------- | --------------------------------------------------------------- | ---------------------- | --------------- |
| **Claude API (Anthropic Compatible)** | Claude models, Chinese models (DeepSeek, Qwen, GLM, Kimi, etc.) | `https://api.lmuai.ai` | Default is fine |
| **OpenAI Responses API Compatible**   | OpenAI models (GPT-5.6 / GPT-5.5, etc.)                         | `https://api.lmuai.ai` | `/responses`    |

<Callout type="info" title="The API Host never includes /v1">
  Chatbox appends the path automatically when it makes a request, so &#x2A;*always enter the API Host up to the domain only — do not include `/v1`**. For example, the OpenAI Responses protocol ends up as `https://api.lmuai.ai/v1/responses`.
</Callout>

***

## Opening the add-provider entry point [#opening-the-add-provider-entry-point]

Both protocols are configured from the same entry point:

<Tabs items="['Desktop', 'Mobile']">
  <Tab value="Desktop">
    Open Chatbox → **Settings in the bottom-left** → go to Model Provider → click Add at the bottom.

        <img alt="Adding a provider in Chatbox desktop" src="__img0" />
  </Tab>

  <Tab value="Mobile">
    Open Chatbox → tap the **menu icon ☰ in the top-left** to open the sidebar → tap Settings → go to Model Provider → tap Add at the bottom.

    Once you're on the configuration screen, the fields map one-to-one to the desktop version (on mobile they're labeled **API Host** and **API Path**); fill them in following Method 1 / Method 2 below.

    <Callout type="info" title="Mobile has its own step-by-step guide">
      The mobile flow — with a **configuration screenshot*&#x2A;, the &#x2A;*"Improve Network Compatibility"** switch, and **checking the assembled address*&#x2A; — is written up as a standalone page: &#x2A;*[Chatbox Mobile](./chatbox-mobile)**.
    </Callout>
  </Tab>
</Tabs>

***

## Method 1: Anthropic protocol (Claude and Chinese models) [#method-1-anthropic-protocol-claude-and-chinese-models]

### Step 1 — Create a provider [#step-1--create-a-provider]

After opening the new-provider window from the entry point above:

1. **API Mode*&#x2A;: select &#x2A;*`Claude API (Anthropic Compatible)`**
2. **Name**: anything you like (e.g. `LMU AI`)

### Step 2 — Enter the API details [#step-2--enter-the-api-details]

* **API Host**: enter the LMU AI API address (**without** `/v1`)

  ```
  https://api.lmuai.ai
  ```

* **API Key**: enter the `sk-` key generated in the console

### Step 3 — Add models [#step-3--add-models]

Chatbox's default list may not include the latest models, so you need to add them manually. On the provider settings page, find the Model section → click New / Add Model and enter the model ID, for example:

* `claude-opus-5`
* `claude-sonnet-5`
* `claude-haiku-4-5`
* `deepseek-v4-pro`
* `qwen3.8-max-preview`
* `glm-5.2`
* `kimi-k3`

For the full list of available models, see the [Model Gallery](/docs/guide/models).

***

## Method 2: OpenAI protocol (GPT models) [#method-2-openai-protocol-gpt-models]

<Callout type="warn" title="Requirements">
  You need to meet one of these conditions:

  * You have subscribed to a **GPT plan**
  * You use pay-as-you-go billing and have selected the **GPT group** in the LMU AI console
</Callout>

### Step 1 — Create a provider [#step-1--create-a-provider-1]

After opening the new-provider window from the entry point above:

1. **API Mode*&#x2A;: select &#x2A;*`OpenAI Responses API Compatible`** (not `OpenAI API Compatible`)
2. **Name**: anything you like (e.g. `LMUAI-GPT`)

### Step 2 — Enter the API details [#step-2--enter-the-api-details-1]

| Field    | Value                                  |
| -------- | -------------------------------------- |
| API Key  | The `sk-` key generated in the console |
| API Host | `https://api.lmuai.ai`                 |
| API Path | `/responses`                           |

Chatbox automatically joins the Host, `/v1`, and Path into the full request address `https://api.lmuai.ai/v1/responses`, and the full assembled URL is shown live below the fields.

Once filled in as above, the screen looks like this:

<img alt="Example of configuring a GPT provider in Chatbox desktop" src="__img1" />

### Step 3 — Add models [#step-3--add-models-1]

In the Model section, click New and enter the model ID:

* `gpt-5.6-sol`
* `gpt-5.6-terra`
* `gpt-5.6-luna`
* `gpt-5.5`
* `gpt-5.4`
* `gpt-5.2`

***

## Usage [#usage]

Back on the main screen, click the model selector at the top, switch to the provider and model you just added, and start chatting.

<Callout type="info" title="One key works with both protocols">
  The LMU AI API key supports both the Anthropic and OpenAI protocols. The API Host is `https://api.lmuai.ai` for both — only the API Mode and API Path differ. If you want to use Claude / Chinese models as well as GPT, just add one provider for each as described above; you can use the same key for both.
</Callout>

***

## Common issues [#common-issues]

### Does the API Host need `/v1`? [#does-the-api-host-need-v1]

**No.** Always enter the Chatbox API Host field up to the domain only: `https://api.lmuai.ai`.

Chatbox appends `/v1` itself:

* Anthropic protocol: becomes `https://api.lmuai.ai/v1/messages`
* OpenAI Responses protocol: set Path to `/responses`, which becomes `https://api.lmuai.ai/v1/responses`

If you add an extra `/v1` in the Host, you get a broken 404 path like `…/v1/v1/...`.

### Why pick OpenAI Responses instead of the plain OpenAI API Compatible? [#why-pick-openai-responses-instead-of-the-plain-openai-api-compatible]

Chatbox's OpenAI API Compatible uses the older `/chat/completions` endpoint; LMU AI's OpenAI protocol uses the newer **Responses API** (`/v1/responses`), so you must pick **OpenAI Responses API Compatible** here and set API Path to `/responses`.

### Chatbox's model list doesn't have Claude Opus 5 / GPT-5.6? [#chatboxs-model-list-doesnt-have-claude-opus-5--gpt-56]

Chatbox's built-in preset model list updates with a delay, so new models need to be added manually via Add Model in the provider configuration — just enter the corresponding model ID. For available models, see the [Model Gallery](/docs/guide/models).

### Changes don't take effect? [#changes-dont-take-effect]

After saving, we recommend **restarting the Chatbox client once**, or reselecting the model in the chat screen.

### 401 / 403 errors? [#401--403-errors]

Check in order:

1. Is the API Key the `sk-` key generated in the LMU AI console (not an official Anthropic / OpenAI key)?
2. Is the API Host wrong (both protocols use `https://api.lmuai.ai`, **neither with** `/v1`; for the OpenAI protocol, also check that API Mode is `OpenAI Responses API Compatible` and API Path is `/responses`)?
3. Does your current plan / billing group include the model you're calling?