Connect ChatGPT (Codex)
Who this is for: anyone with a ChatGPT plan that includes Codex who wants to use that plan through SpiderGate.
What this does: walks you through signing in with ChatGPT from the SpiderGate vault, choosing your plan package, and keeping the login to your own brand.
Before you begin
A ChatGPT plan that includes Codex. OpenAI says Codex is included in ChatGPT Free, Go, Plus, Pro, Business, Edu and Enterprise.
A browser where you are signed in to that ChatGPT account.
About 2 minutes.
Limits and terms to know first
Your plan's limits apply, not API billing. OpenAI publishes estimates, not caps: Plus gets a number of messages per five hours that depends on the model, Pro "currently" has no five-hour limit, and "Weekly limits may also apply". ChatGPT and Codex share one allowance, so work you do in ChatGPT counts too. OpenAI retires models from Codex sign-in on its own schedule: GPT-5.5 left on 14 October 2026.
Warning: OpenAI says "API keys are still the recommended default for automation", and offers ChatGPT-plan use in other companies' apps through its own "Sign in with ChatGPT" programme, which is open to selected partners. For agents and scheduled jobs, an OpenAI API key is the path OpenAI recommends.
Warning: For its own CI use of ChatGPT sign-in, OpenAI says to use "one auth.json per runner" and not to share it "across concurrent jobs or multiple machines". Keep this login to your own brand: switch off pool sharing in Step 4.
Steps
Step 1 — Start the connection
Open Gate → Vault (/dashboard/gate/vault) and click Add Provider. Under Connect a subscription, choose Codex ("Sign in with ChatGPT"). The Connect Codex dialog opens.
Step 2 — Pick your plan package
Under Billing · Subscription, pick the package that matches your ChatGPT plan. Billing is fixed to Subscription for a login. The package tells SpiderGate your five-hour and weekly allowance; see Subscription keys & package tiers. Optionally set CLI leases, the number of runners that may borrow this login at once.
Step 3 — Sign in and paste back
Under Step 1 — Sign in, click Open ChatGPT sign-in. In the new tab, sign in to your ChatGPT account and authorize Codex. You land on a localhost:1455 page that does not load. That is expected.
Copy the full URL from your browser's address bar. Back in SpiderGate, paste it into Step 2 — Paste the full URL from your browser's address bar, and click Finish connect.
Step 4 — Keep the login private
A Codex login is shared with the SpiderIQ key pool by default. Open the new key in your vault, choose Edit, and turn off Share with SpiderIQ key pool.
Verify it worked
The login appears in your vault as a Subscription key, with ChatGPT plan: …, from OpenAI next to the package: that is the plan OpenAI reports for the login. If it does not match the package you chose, the key card warns you.
Ask the gateway what you can use with GET /api/gate/v1/subscriptions (see Your Subscriptions). A Codex subscription comes back with usage: completion and a list of model ids you can pin.
Troubleshoot
Couldn't read a code + state from that. Paste the full URL or code shown by the provider. You pasted part of the URL. Copy the whole address, starting http://localhost:1455/auth/callback?code=.
The sign-in tab never redirects. Sign in to ChatGPT in that browser first, then click Open ChatGPT sign-in again.
Requests fail once you hit your plan's limit. That is your ChatGPT allowance, shared with your own use of ChatGPT. Check your usage in ChatGPT; it resets on OpenAI's schedule.