Use ChatGPT Plus / Pro as a Coding Agent
If you're already paying $20/mo (Plus) or $200/mo (Pro) for ChatGPT, that subscription can power a real coding agent on VibeKit, via the same Codex OAuth flow OpenAI's official Codex CLI uses. No second AI subscription. No per-token charges on top.
The math
Buying OpenAI API tokens for ChatGPT-tier usage gets expensive fast: a typical coding session of 10-20 agent prompts can run $1-$5 in API tokens. ChatGPT Plus ($20/mo) covers that volume daily as part of the flat subscription, with substantial quota left over. ChatGPT Pro ($200/mo) is effectively unlimited for individual use.
OpenAI made Codex OAuth specifically so subscribers can use their plan for coding-style automation without spinning up a separate Developer-account billing relationship. VibeKit supports the same auth path.
What you actually get
- Same model OpenAI uses for Codex, currently the GPT-5 family. Same quality, same context window.
- Same usage caps as ChatGPT Plus / Pro directly, not a separate per-token meter on top.
- Zero VibeKit markup. BYOK is BYOK. We don't bill anything on the AI side.
- Real container hosting included. Free tier covers 2 workspaces, with apps on Fargate.
Setup
- Install VibeKit (App Store) or visit app.vibekit.bot
- Settings → AI Provider → Sign in with ChatGPT
- OpenAI's OAuth screen confirms your Plus / Pro subscription
- Pick a template or connect a GitHub repo
- Start chatting, agent calls now run against your existing subscription
How this stacks against the alternatives
| Option | Monthly cost | Best for |
|---|---|---|
| ChatGPT Plus + Codex OAuth (VibeKit) | $20 (existing sub) | You already pay for Plus |
| ChatGPT Pro + Codex OAuth (VibeKit) | $200 (existing sub) | Heavy daily coding |
OpenAI sk- API key | Pay per token | Light or bursty usage |
| Platform fallback (no BYOK) | OpenRouter price + 20% | Trying VibeKit out |
| Free pool on VibeKit | $0 | Everyday tasks, prototypes |
You can mix Claude and ChatGPT per workspace
Connect both subscriptions in Settings → AI Provider (each connection covers your whole account), then pick each workspace's model in its model picker. Run your marketing site on a ChatGPT model through Codex OAuth and your serious DeFi app on Claude through your Pro plan. Changing one workspace's model leaves the others alone.
FAQ
What is Codex OAuth?
Codex OAuth is the authentication flow OpenAI's official Codex CLI uses to let ChatGPT Plus and Pro subscribers run a coding agent against their existing subscription quota, no API key, no per-token billing. VibeKit supports the same OAuth flow: sign in with your ChatGPT account in Settings → AI Provider and your coding agent runs against your existing $20 (Plus) or $200 (Pro) per month subscription.
Does this mean I get unlimited coding for $20/mo?
It means you get the same usage allowance OpenAI gives Plus subscribers on Codex, which is generous but not unlimited. Pro ($200/mo) has dramatically higher limits. Either way, you don't pay anything extra to VibeKit on the AI side; we run zero markup on BYOK. Our paid plans cover hosting + more workspaces + custom domains.
How does this compare to using an OpenAI API key (sk-)?
API key is pay-per-token. Codex OAuth is flat-rate subscription. For a developer doing 10+ agent tasks a day, Codex OAuth via ChatGPT Plus often comes out cheaper than the equivalent in API tokens. For light usage or bursty work, the API key might be cheaper. Your account holds one OpenAI connection at a time, so you can switch between them any time in Settings → AI Provider and every workspace follows.
Does it work the same as Claude Pro OAuth?
Yes, same idea, different provider. Claude Pro / Max OAuth uses your Anthropic subscription quota; ChatGPT Plus / Pro Codex OAuth uses your OpenAI subscription quota. Both are zero-markup BYOK paths on VibeKit. Pick whichever subscription you already have, or use both on different workspaces.