Choosing a backend

Ropenode drives two agent CLIs. You choose one when you create a session, from the same picker as the model.

	Claude Code	Codex
Binary	claude	codex
Sign-in	claude on first run	codex login
Credential lives in	the CLI's own store	~/.codex/auth.json
Models	fable, opus, sonnet, haiku	whatever ~/.codex/config.toml says
Edit policy, diff review, undo, Log	yes	no — see below
MCP servers from Ropenode	yes	no — Codex reads its own config

The one rule both obey

Every session is driven by a CLI that owns its own authentication — the app never holds a model credential.

There is no code path that could store your key, because nothing in Ropenode talks to a model API.

Picking a model

For Claude, pick a model and Ropenode passes it to the CLI.

For Codex, Ropenode deliberately passes no --model flag at all, deferring to your ~/.codex/config.toml — your config is more likely to be right than our guess.

If a model does not belong to the session's agent, Ropenode leaves the choice to the CLI and tells you.

Using both in one project

The agent is fixed when a session is created, so to hand the same work to the other agent, start a new session in the same project. A project can hold both kinds side by side.

Which should you use?

Whichever you already pay for. If you have both, the practical difference today is supervision coverage — read the next page before deciding.