I Built an Open-Source Telegram Bot for the OpenAI Codex Referral Promo

A developer built an open-source Telegram bot called @codexHuddbot to solve the one-shot referral matching problem for OpenAI's Codex referral promo. The bot stores referral links in a queue and matches them one-to-one on demand, preventing wasted referrals. The project is MIT-licensed and available on GitHub.

OpenAI dropped Codex referral links on June 12. Seven days of free Codex Pro for you, seven days for a friend. The catch? Each referral code can only be used once. Post your link on social media and the first person to click it claims it — everyone else gets "this referral has already been used." One shot, gone. That's a poor user experience. And it's wasteful: every unused referral is a missed opportunity for someone who actually wants to try Codex. So I built @codexHuddbot — a Telegram bot that solves the one-shot matching problem. A user sends their Codex referral link to the bot. The bot stores it in a queue. When another user asks for a referral, the bot dequeues the oldest link, marks it as claimed, and sends it privately. No public link dumps. No race conditions. No wasted referrals. One-to-one matching, on demand. The bot handles duplicates you can't submit the same link twice , tracks which links have been consumed, and provides a /status command. I wanted something that could ship in an afternoon and run reliably for the two-week promo window. pip install aiohttp and systemctl start codex-buddy . This is the live, running component. wham/rate-limit-reset-credits/consume endpoint. For Linux and headless users who never see the in-editor popup — they'd otherwise leave those credits unused.~500 lines of application code. Two repos, MIT-licensed. The promo ends June 24. After that, the bot's core function retires. But the architecture — a lightweight Python aiohttp server behind nginx, backed by SQLite — is reusable for any time-limited Telegram bot. The code is clean enough to serve as a reference, and the referral-matching logic is a simple queue pattern you can adapt. Both MIT-licensed. Stars, issues, and PRs welcome. The promo ends June 24 — if you've got an unused referral link, now's the time to share it.