Codex with a custom model, in one command

No ChatGPT account needed; the model catalog comes with the command, so long tasks use each model’s real context length.

Your key is filled in when you copy
bash <(curl -fsSL 'https://api.tutucompute.com/v1/launch/codex?lang=en' -K - <<< 'header = "Authorization: Bearer …"')

Get started in three steps

  1. 1

    Copy the command on this page (it comes with a free key, no sign-up)

  2. 2

    Paste it into a terminal and press Enter

  3. 3

    Codex is installed if missing and opens

How the switch works

Codex’s own launch flags (-c model_provider=…) point it at TutuCompute — config.toml is not written. The model catalog lives in ~/.codex/tutu/, and plain codex keeps your usual setup.

Use the full context window

With a model catalog Codex knows each model’s real context length. If your Codex version doesn’t accept it, it still starts and tells you so.

Rather not run a script?

The connect page also gives native settings: one environment variable per line plus a launch command to paste into your terminal.

FAQ

Related