Skip to main content
The ChatGPT desktop app and the Codex CLI both read ~/.codex/config.toml, so connecting one connects the other. Quit the ChatGPT desktop app and close open Codex sessions before connecting or running off. The app loads its model list only at startup. Nexus CLI waits for both to close, or use --force to write anyway. Generally available OpenAI Responses Nexus Auto Native web search macOS only

​
Quick start

Install and sign in

Requires Node.js 18 or later, git, and bash. On Windows, run it from Git Bash; PowerShell breaks the script’s line endings.
Sign in with your browser, or paste a Fireworks key.Already set up? Skip to step 2.

Connect the ChatGPT desktop app

nexus chatgpt is the same command as nexus codex.

Reopen the app

Reopen the app. Its model list starts on Nexus Auto (auto) and adds the Fireworks models next to the OpenAI models.
The model list starts on Nexus Auto (auto) and shows the Fireworks models.

​
Use Nexus Auto

The app keeps its own ChatGPT sign-in, which pays for OpenAI model turns. Your Fireworks key pays for Fireworks models. See Who pays for OpenAI model turns.

​
MCP and tools

MCP servers and plugins keep working, through Codex’s tool search. Native web search is supported while connected through Nexus CLI. Fireworks runs the search server-side and returns the results, with no MCP server or search service to set up.

​
Troubleshooting

Upgrading from an earlier version? See Upgrading from earlier versions.
Threads keep the provider they started on. A thread started on an OpenAI model before you connected stays on OpenAI, even after you pick Nexus Auto or a Fireworks model. ChatGPT then shows an error such as The 'auto' model is not supported when using Codex with a ChatGPT account. The error names whichever Fireworks model you picked.
  • Start a new thread to use Nexus Auto and Fireworks models.
  • To keep working in the old thread, switch back to an OpenAI model, or run nexus chatgpt off and turn it back on when you start a new thread.
  • After off, threads started on Fireworks models don’t open until you run nexus chatgpt again.
Sign-in billing is tested in the Codex CLI. The ChatGPT desktop app is not tested yet.

​
Manual setup (advanced)

Quit the app, follow the Codex manual setup, then reopen the app and pick the model you set.

​
What the Nexus CLI writes

Nexus CLI writes the same config.toml and model catalog as for Codex; see What the Nexus CLI writes.