March 15, 2026
Free lunch or ban brunch?
Show HN: Free OpenAI API Access with ChatGPT Account
Free AI via your ChatGPT login? Hype vs ban-hammer panic
TLDR: A community tool claims you can use OpenAI through your ChatGPT login without extra fees. Commenters are split between “ban incoming” warnings and hopes it becomes official, with ToS worries and buffet memes highlighting the risk of building anything serious on it.
Developers are buzzing over a new community tool that claims “free OpenAI API” via your ChatGPT login — no extra key, just your existing account routed through a small local proxy. It’s unofficial, uses your own sign‑in tokens, and loudly warns: personal use only. The reaction? Half the thread cheering the free lunch, the other half bracing for the ban hammer. User redgridtactical predicts a “short shelf life” before OpenAI notices unusual traffic and shuts it down. Others say this looks like the long‑rumored “Sign in with OpenAI,” with mrtksn wondering if it’ll be banned or blessed. EvanZhouDev even dropped receipts of OpenCode “collaborating” on ChatGPT login support, fanning the “is this sanctioned?” flames.
Optimists like skeptrune think OpenAI will “make this official rather than ban,” but the caution crew stole the show. TZubiri’s ToS riff went viral: it’s like hitting an all‑you‑can‑eat buffet, bringing your whole extended family, and trying to pay once — hilarious, and also the ethics/legal red flag. Builders warn not to base startups on this; if OpenAI patches it, your app craters. Memes about “ban speedrun%,” “stuffing Tupperware at the buffet,” and “free‑tier heist” flew, but beneath the jokes is a real question: will personal ChatGPT plans power third‑party apps, or will OpenAI slam the door?
Key Points
- •The openai-oauth project exposes a local OpenAI-compatible endpoint at http://127.0.0.1:10531/v1 using ChatGPT/Codex OAuth tokens, requiring no API key.
- •Two usage modes are provided: a CLI for a localhost proxy and a Vercel AI SDK provider via createOpenAIOAuth.
- •Supported endpoints include /v1/responses, /v1/chat/completions, and /v1/models, with streaming, tool calls, and reasoning traces enabled.
- •Configuration options cover host, port, model allowlists, Codex API version, base URL, OAuth client ID, token URL, auth file paths, and automatic token refresh.
- •The project is unofficial and unaffiliated with OpenAI, intended for personal, local experimentation only, with users responsible for compliance with OpenAI terms.