How to connect your own Claude or ChatGPT account
Este artigo ainda não está traduzido, por isso aparece em inglês.
Sign in to Claude Code or Codex with your own account in the sandbox terminal. It takes about a minute, and your sign-in is deleted when you finish.
Some coding screens and hackathons let you work with your own AI agent. If yours does, the start page says so before you begin, and the editor's header reads "AI assistant: run claude in the terminal and sign in with your own account". You run the agent in the sandbox terminal and sign in to your own Claude or ChatGPT account. Nothing to install.
What you need
- For Claude Code: a Claude account on a plan that includes Claude Code.
- For Codex: a ChatGPT account on a plan that includes Codex.
Usage counts toward your own plan's limits. Neither Kendor nor the organiser pays for it or can raise it, so check your limits before a long session like a hackathon.
Sign in to Claude Code
- Open a terminal in the editor and type
claude, then press Enter. - The terminal prints "Sign in with your own Claude account: open the link Claude Code shows, then paste the code back here."
- When Claude Code asks how to sign in, choose your Claude account (the subscription option).
- Open the link it shows. If clicking doesn't open it, copy the link from the terminal into a new browser tab.
- Sign in to Claude and approve the request. Claude shows a code.
- Copy the code, paste it into the terminal and press Enter.
Claude Code then starts in your workspace. Next time you type claude, it opens straight away.
Sign in to Codex
- Open a terminal and type
codex, then press Enter. - Codex prints a link and a one-time code.
- Open the link, sign in to ChatGPT and enter the code.
Codex starts as soon as you've approved it. codex login does the same thing. Codex uses a code instead of sending you back to the browser because the sandbox runs in the cloud, not on your computer.
How long the sign-in lasts
- Across challenges: when you submit one challenge and open the next, your sign-in comes with you.
- When you finish: your sign-in is deleted when you submit the whole assessment or your time runs out.
- Sooner: type
/logoutinside Claude Code, or runcodex logout.
Who sees what
- Your password and sign-in: nobody. You sign in on Anthropic's or OpenAI's own page, and the agent talks to them directly, not through Kendor.
- Your conversation with the agent: the organiser's reviewers or judges see your prompts, the agent's replies and the commands it ran, along with your terminal and code. That's how they see how you work.
- Your files and prompts: they go to Anthropic or OpenAI under your account, on that account's terms, as they would on your own computer.
If something goes wrong
| What you see | What to do |
|---|---|
| "The AI assistant is not enabled for this assessment." | The organiser hasn't turned agents on. Work without one, or ask them. |
| "codex is not enabled for this assessment — use claude." | Only one agent is allowed. Use the one it names. |
| The link doesn't open | Copy it from the terminal into a new tab. |
| Claude or ChatGPT says you've hit a usage limit | It's your plan's limit. Wait for it to reset, or carry on without the agent. |
| You're asked to sign in again after a reconnect | Your sandbox restarted. Sign in again; your work is still there. |
| Your company's account won't sign in | Admins sometimes block sign-ins from new places. Try a personal account. |
Tip. If your hackathon gives you a warm-up sandbox before the day, sign in there first, so you know it works before your clock starts.