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

# Codex

> Use WaveAPI text models in Codex CLI and the Codex desktop app

Add WaveAPI as a model provider in Codex's `config.toml` to use GPT, Claude, Gemini, DeepSeek, GLM, Kimi, Qwen and other text models in Codex. Codex CLI and the Codex desktop app read the same configuration.

## Prerequisites

* Codex CLI installed (`codex --version` prints a version) or the Codex desktop app. See the [Codex docs](https://developers.openai.com/codex) for installation.
* An API key created in the [console](https://wave.qingbo.ai/dashboard/keys).

## Configure

### 1. Set the API key

Codex reads the key from the `WAVEAPI_API_KEY` environment variable.

<CodeGroup>
  ```zsh macOS theme={"system"}
  read -s "WAVEAPI_API_KEY?WaveAPI API Key: "; echo
  echo "export WAVEAPI_API_KEY=\"$WAVEAPI_API_KEY\"" >> ~/.zshrc
  launchctl setenv WAVEAPI_API_KEY "$WAVEAPI_API_KEY"
  ```

  ```bash Linux theme={"system"}
  read -rs -p "WaveAPI API Key: " WAVEAPI_API_KEY; echo
  echo "export WAVEAPI_API_KEY=\"$WAVEAPI_API_KEY\"" >> ~/.bashrc
  ```

  ```powershell Windows (PowerShell) theme={"system"}
  $key = Read-Host "WaveAPI API Key" -AsSecureString
  [Environment]::SetEnvironmentVariable("WAVEAPI_API_KEY", [System.Net.NetworkCredential]::new("", $key).Password, "User")
  ```
</CodeGroup>

`launchctl setenv` makes the variable visible to the Codex desktop app; restart the app afterwards. On Windows, open a new terminal after setting it.

### 2. Edit config.toml

Edit `~/.codex/config.toml` (Windows: `%USERPROFILE%\.codex\config.toml`):

```toml theme={"system"}
model = "gpt-5.4"
model_provider = "waveapi"

[model_providers.waveapi]
name = "WaveAPI"
base_url = "https://api.qingbo.ai/v1"
env_key = "WAVEAPI_API_KEY"
wire_api = "responses"
```

| Field            | Value                                                 |
| ---------------- | ----------------------------------------------------- |
| `model`          | Default model ID                                      |
| `model_provider` | `waveapi`, matching `[model_providers.waveapi]` below |
| `base_url`       | `https://api.qingbo.ai/v1`                            |
| `env_key`        | Name of the environment variable holding the key      |
| `wire_api`       | `responses`                                           |

<Warning>
  `model` and `model_provider` must be at the top of the file, before any `[...]` table. Inside a table they become fields of that table, and Codex keeps asking you to sign in to ChatGPT.
</Warning>

## Verify

```bash theme={"system"}
codex exec "Introduce yourself in one sentence."
```

A reply means the configuration works.

## Available models

Every text model except `deepseek-v3.1-terminus` can be used in Codex. Model IDs and prices are in the [Text models overview](/en/api-reference/text/overview).

Switch models with `/model` in Codex, or pick one at launch:

```bash theme={"system"}
codex -m claude-sonnet-5
```

## Generate images and videos

Codex can also generate images, videos and audio through WaveAPI's MCP tools; see [MCP & Skill](/en/integrations/agent-tools).

## Common errors

| Symptom                                                    | Cause                                             | Fix                                                                                                               |
| ---------------------------------------------------------- | ------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------- |
| Asks you to sign in to ChatGPT on start                    | `model_provider` did not take effect              | Make sure `model_provider = "waveapi"` is at the top of the file, before any `[...]` table                        |
| Says the `WAVEAPI_API_KEY` environment variable is missing | The Codex process cannot see the variable         | Terminal: open a new terminal or run `source ~/.zshrc`. Desktop app: run `launchctl setenv`, then restart the app |
| `401`                                                      | The API key is invalid or disabled                | Check the key in the console                                                                                      |
| `400` saying the model is unavailable                      | The model does not support the Responses endpoint | Use any model other than `deepseek-v3.1-terminus`                                                                 |