Two Codex accounts.
One AgentsRoom. Zero shell tricks.

There is no hard limit. Each account is a directory on disk. Add a Work account, a Personal account, one account per client, a test account: AgentsRoom lists them all in the Settings panel and lets you pick any of them per project or per agent.

No. AgentsRoom drives Codex CLI's CODEX_HOME mechanism directly. No external switcher, no shell wrapper, no extra dependency. The sign-in flow is fully embedded in the Settings panel and runs the official codex login command.

Yes. The project pins a default account, but each agent can override it. Open a project pinned on the Work account, edit one specific agent to use the Personal account, and the two agents run in parallel on different Codex accounts in the same window.

Nothing breaks. The agents fall back to the default Codex account silently. The deleted directory is left on disk in case you want to point another profile at it later. You can rebind the orphaned agents to another account at any time from the Edit Agent dialog.

No. Each account directory stays on your local machine. AgentsRoom never reads, copies or transmits credentials. The sign-in flow runs the official codex login CLI in a local PTY, OAuth happens between OpenAI and your browser as usual.

Yes. The Multiple Accounts panel in Settings handles both Claude and Codex accounts independently. A single project can have one Claude account pinned and one Codex account pinned. Each agent uses whichever provider and account it is configured for.

Yes. Sidebar folders carry their own defaults, and the Codex account is one of them. Pin Work on the folder holding your work repositories and every project inside starts on that account, including projects added later and projects in sub-folders. The resolution order is: agent override, project pin, nearest folder that pins an account, global default, then ~/.codex.

The one you are actually spending, resolved separately for each provider: the account of the agent in the foreground first, otherwise the account this project would launch on, otherwise your primary account. The gauge tooltip names it, and the usage popover lists one section per signed-in account, so a work quota and a personal quota are never added into a single figure.

Yes, with the quota breakdown enabled in the beta panel. AgentsRoom takes the percentage published by the account for the running window and splits it across your projects and agents, weighted by the local cost of each session. It is an estimate and it is marked as such: consumption AgentsRoom cannot see, from another computer or from another tool, is missing from the calculation, so this is never an invoice.

Turn on "Use my personal CLI configuration" in Settings and yes. AgentsRoom runs Codex against a generated CODEX_HOME per project, which is what lets several accounts and several projects run side by side without colliding: Codex then reads its user layer from that directory, so your own ~/.codex prompts, skills and global AGENTS.md are not loaded until you share them. The switch links them into every generated home and keeps them up to date, while credentials, sessions and history stay isolated per account.

You do not log out at all, and that is the point. Every account you add keeps its own signed-in state on disk, so switching is picking a different account from a dropdown rather than running codex logout and codex login again. Three places do the switching, from the widest to the narrowest: the global default in Settings, the account pinned on a project, and the override on one agent. Change any of them and the next agent you start runs on that account, while agents already running keep the account they were launched with. Nothing is re-authenticated, no browser opens, and no session is destroyed, which is exactly what a logout and login cycle would cost you every time. The account switcher also runs by itself if you want: with automatic switching on, an agent that hits its account's usage limit moves to the next account you signed in instead of stopping.