OpenAI Codex

If you have a ChatGPT Plus or ChatGPT Pro subscription, you can connect it directly to Evonic. No API keys, no per-token billing — just your existing subscription.

How It Works

Instead of the usual API key approach, Codex uses a secure OAuth 2.0 + PKCE flow. You’ll set up Codex as a provider just like any other model provider — the only difference is that you authorize it through your OpenAI account instead of pasting an API key.

Here’s the overall flow:

1. Create a provider — add a new provider with API Format set to “Codex (OAuth)”
2. Connect — click Connect on the provider card to authorize via your OpenAI account
3. Fetch models — after connecting, fetch the list of available models from your subscription

Your data still flows through Evonic — OpenAI never sees it directly. The subscription gives you access to models like gpt-5.6-sol, gpt-5.6-terra, and gpt-5.6-luna with no per-token charges.

Prerequisites

• An active ChatGPT Plus ($20/month) or ChatGPT Pro ($200/month) subscription
• Evonic running and accessible in your browser
• Internet connection (authentication requires reaching auth.openai.com)

Connecting Your Subscription

Setting up Codex is a three-step process: create the provider, authorize it, then fetch your models.

Step 1 — Create a Provider

Navigate to Settings → Models and click Add Provider. Fill in the provider form:

Field	Value
Display Name	OpenAI Codex (or any name you prefer)
Type	Remote
Base URL	https://chatgpt.com/backend-api/codex
API Key	Leave empty — OAuth will fill this in
API Format	Codex (OAuth)

Click Save Provider. A new provider card appears in the models grid with Connect and Fetch Models buttons.

Step 2 — Connect & Authorize

On the provider card, click the green Connect button. A popup window opens, redirecting you to OpenAI’s login page. Log in with your OpenAI account and approve the connection. OpenAI will ask you to confirm that Evonic can access your subscription.

Once authorized, Evonic exchanges the authorization code for secure tokens automatically. The provider card shows a green dot when connected.

Remote access?

If you’re running Evonic on a remote server, the callback popup may not work. Paste the full callback URL from your browser address bar into the text area in the waiting modal.

Step 3 — Fetch Models

After connecting, click the Fetch Models button on the provider card. Evonic queries the Codex endpoint and lists all models available under your subscription. From the results, you can quickly add any model with one click.

That’s it! Your Codex models are now ready to use — assign them to any agent like you would with any other model provider.

Available Models

The models available depend on your subscription tier. After clicking Fetch Models, you’ll see a list of what’s available:

Subscription	Typical Models
ChatGPT Plus	gpt-5.6-sol, gpt-5.6-terra
ChatGPT Pro	gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna, plus extended context windows

If the API doesn’t return a model list, Evonic falls back to these defaults so you can still add them manually.

Using Codex with Agents

Once connected and models are fetched, Codex models appear in the model dropdown when you create or edit an agent. Here’s how it looks in your configuration:

model:
  provider: openai-codex
  model_name: gpt-5.6-sol

The provider field matches the provider name you chose when creating it. The model_name is the model ID shown in the Fetch Models result.

From the CLI, you can verify the connection:

evonic model list

Example output:

ID                                    Name           Provider       Status
--------------------------------------------------------------------------
a1b2c3d4-...                         GPT-5.6 Sol     openai-codex   enabled
e5f6g7h8-...                         GPT-5.6 Terra   openai-codex   enabled

Managing Your Connection

Check Connection Status

The Codex card in Settings → Models shows a green indicator when connected. Click it to see token expiry details.

Reconnecting

Tokens expire periodically. Evonic refreshes them automatically in the background. If the refresh fails — for example, if you changed your OpenAI password — just click Connect again to re-authorize.

Disconnecting

To remove the connection, click Disconnect on the Codex card. This deletes all stored tokens and removes Codex models from your agent configuration. You can reconnect at any time.

How It Differs from Standard OpenAI API

Feature	Standard API	Codex (Subscription)
Authentication	API key (sk-...)	OAuth 2.0 login
Billing	Per-token	Included in subscription
Endpoint	api.openai.com/v1	chatgpt.com/backend-api/codex
API format	Chat Completions	Responses API
Streaming	SSE	WebSocket (SSE fallback)
Setup	Paste API key	One-click OAuth

For most users, Codex is the simpler option — no need to manage API keys or monitor token usage.

Troubleshooting

”Connection Failed” after clicking Connect

• Make sure your browser allows popups from Evonic
• Check that auth.openai.com is reachable from your network
• Verify your ChatGPT subscription is active at chatgpt.com/account

Models not appearing after authorization

• Make sure you clicked Fetch Models after connecting — models are not auto-discovered
• Check that your subscription is active at chatgpt.com/account
• If the API returns an empty list, try adding models manually using the fallback names (e.g., gpt-5.6-sol)

“Token Expired” warning

• Tokens refresh automatically; the warning appears when the refresh window is closing
• If the warning persists for more than a few minutes, re-authorize by clicking Connect again

Slow responses compared to API

Codex uses WebSocket transport with SSE fallback. If responses feel slow:

• Ensure your network connection is stable
• Check that WebSocket connections aren’t blocked by a firewall or proxy
• The first request after a period of inactivity may have slightly higher latency