# CC Switch Setup

[CC Switch](https://ccswitch.io/) is a cross-platform configuration manager for AI CLIs. It provides a visual way to switch between the official Codex setup and multiple API providers. This guide uses **CC Switch 3.20 or later**.

## Configuration Overview

CC Switch manages `~/.codex/auth.json` and `~/.codex/config.toml`:

| Field | Value |
| --- | --- |
| Provider name | `4AICode` |
| API Key | An `sk-4ai-...` token from the dashboard |
| API endpoint | `https://relay.4aicode.com/v1` |
| Default model | `gpt-5.6-terra` |
| Upstream format | `OpenAI Responses API` |

4AICode supports the Responses API directly. Keep Full URL mode and local routing disabled.

## 1. Prepare the Token and Clients

1. Create an `sk-4ai-...` token from [Dashboard → Manage Tokens](https://4aicode.com/tokens).
2. Install the latest version from the [CC Switch website](https://ccswitch.io/) or [GitHub Releases](https://github.com/farion1231/cc-switch/releases/latest).
3. Install the [Codex CLI](https://developers.openai.com/codex/).

`sk-4ai-xxxxxxxxxxxxxxxx` is a placeholder in this guide. Replace it with the token created in the dashboard when filling in the provider. Keep real tokens out of version control and private messages.

On first launch, CC Switch can import your existing Codex configuration. It updates the Codex configuration files whenever you switch providers.

## 2. Add the 4AICode Provider

1. Select **Codex** at the top of CC Switch.
2. Click **+** in the upper-right corner.
3. Stay on **App-specific provider** and select the **Custom** preset.
4. Enter the following values:

| CC Switch field | Value |
| --- | --- |
| Name | `4AICode` |
| API Key | `sk-4ai-xxxxxxxxxxxxxxxx` |
| API endpoint / Base URL | `https://relay.4aicode.com/v1` |
| Default model | `gpt-5.6-terra` |

The recommended endpoint depends on region: use `.com` on international networks and `https://relay.4aicode.net/v1` in mainland China. The English examples use `.com`.

## 3. Check Advanced Options

Expand **Advanced options** and confirm:

| Option | Setting |
| --- | --- |
| Upstream format | `OpenAI Responses API` |
| Full URL mode | Off |
| Local routing / mapping | Off; recent CC Switch versions may infer this from the upstream format and hide the switch |
| Model mapping | Empty |

`https://relay.4aicode.com/v1` is an API prefix. With Full URL mode disabled, Codex adds the `/responses` path.

In the **config.toml (TOML)** editor near the bottom of the form, set `model_reasoning_effort` to `xhigh`. This is the recommended value for this guide, and CC Switch writes it to the Codex configuration.

Click **Add** to save the provider.

## 4. Enable and Verify

1. Click **Enable** on the 4AICode provider card.
2. Quit every running Codex CLI or Codex App instance.
3. Open a new terminal and run:

```bash
codex
```

Enter a simple task such as “List this directory and explain the project structure.” A normal response confirms the setup.

Codex needs a restart after CC Switch changes its provider. You can use the main window or system tray for future provider switches.

## Change Models

Edit the 4AICode provider and enter a model under **Default model**. These models are currently exposed by 4AICode; the dashboard model list and account-pool status are authoritative:

| Model | Recommended use |
| --- | --- |
| `gpt-6-astra` | New-generation flagship; billed at 15% of the official price |
| `gpt-5.6-sol` | Complex engineering and strongest reasoning |
| `gpt-5.6-terra` | Everyday coding; recommended default |
| `gpt-5.6-luna` | Fast, economical tasks |
| `gpt-5.5`, `gpt-5.4`, `gpt-5.4-mini`, `gpt-5.3-codex` | Existing workflows |

`gpt-5.6-sol` supports `max` reasoning effort. `xhigh` is the recommended setting for the other models. `max` uses more reasoning compute, responds more slowly, and costs more. Save, enable the provider again, and restart Codex.

## Configuration Reference

CC Switch stores the token in `auth.json` for a third-party API-key provider, with the same semantics as the examples below. The official ChatGPT login mode uses a separate authentication structure containing `tokens` and `last_refresh`; CC Switch manages each structure according to the provider type. CC Switch may also generate a different provider ID and section name.

`auth.json`:

```json
{
  "OPENAI_API_KEY": "sk-4ai-xxxxxxxxxxxxxxxx"
}
```

`config.toml`:

```toml
model_provider = "custom"
model = "gpt-5.6-terra"
model_reasoning_effort = "xhigh"

[model_providers.custom]
name = "4AICode"
base_url = "https://relay.4aicode.com/v1"
wire_api = "responses"
requires_openai_auth = true
```

`requires_openai_auth = true` tells this provider to read its API key from `auth.json`.

## Troubleshooting

| Symptom | Fix |
| --- | --- |
| Codex still uses the previous provider | Confirm the 4AICode card says “Currently enabled,” quit Codex completely, then open a new terminal or restart the app |
| A request returns `401` | Check that the API key is complete and active in the dashboard |
| A request returns `404` | Ensure the Base URL ends in `/v1` and Full URL mode is off |
| Requests go through the CC Switch local proxy | Select the Responses format and disable local routing takeover |
| The fields in this guide are missing | Upgrade CC Switch; field labels follow the current app version |

For the setup script, Codex App mobile support, and SDK examples, see [ChatGPT / Codex Setup](/docs/en/guides/chatgpt).