Run Codex with your whole team, in your own cloud

Published July 19, 2026 · by the AQ team

AQ runs Codex, OpenAI's coding agent CLI, in cloud workspaces your whole team shares: the same live terminal, code editor, and app preview in everyone's browser. Codex runs in tmux on a VM you control, and each workspace gets its own isolated git worktree, so runs survive closed laptops and parallel workspaces never collide. You sign in with the ChatGPT plan or OpenAI account you already have; AQ never proxies or marks up your model usage.

OpenAI also offers cloud Codex tasks inside ChatGPT. The difference in AQ is control and company: execution happens on your own infrastructure rather than a vendor sandbox, the session is a real terminal you can watch and interrupt rather than a task that returns a diff, and the workspace is multiplayer, so teammates review the work while it happens instead of after.

What running Codex in AQ adds

Sessions that outlive your laptop. Codex runs in tmux on the workspace VM. Close the lid or lose Wi-Fi and the run continues; the scrollback is intact when you return, from any device.
Parallel Codex runs, zero collisions. Every workspace is an isolated git worktree on the same repository, so many Codex runs, one per workspace, can work many tasks at once. (Background: git worktrees for AI coding agents.)
A terminal your team can enter. Teammates open the workspace URL and see the same live session: they can watch Codex work, type into it, or take it over. No screen-share, no pairing tool, nothing to install.
Previews with a feedback loop. The workspace serves its dev server behind a shareable link, guest links included, and comments pinned on the preview can be sent to the agent as prompts.
Tracker-driven intake. Label a Linear issue "ai-task" and a workspace appears in the team sidebar, with status syncing back as work progresses.
PRs tracked per workspace. Codex commits and opens pull requests with your own GitHub sign-in; AQ records opened, merged, and closed PRs against the workspace.

How setup works

Connect Codex in AQ's settings with the standard interactive sign-in, once per user, on whatever ChatGPT or OpenAI plan you already pay for. Connect a repository and choose where agents run: your own VM on the Free plan, or a dedicated AQ-managed VM on the Team plan. Then create a workspace and start Codex from its terminal. Teammates who open the workspace join the same live session.

What AQ deliberately does not change

The Codex inside an AQ workspace is the unmodified CLI with your own login, running in a real terminal. New Codex features arrive when OpenAI ships them, not when AQ catches up. And the workflow is portable by construction: git worktrees, tmux, and a stock CLI port out of AQ as cleanly as they ported in.

Codex alongside Claude Code and every other agent

AQ runs Codex next to Claude Code, Cursor Agent, Kimi, Grok, Pi, OpenCode, and plain shells, in separate workspaces on the same repository. Teams use that to compare agents on the same task in parallel worktrees, and to split work by strength across CLIs. The same story from the other side: running Claude Code with your team.

Frequently asked questions

Do I need an OpenAI API key to run Codex in AQ?

No. You sign in with your existing ChatGPT or OpenAI account through Codex's normal login flow, once per user. AQ does not sit between the CLI and OpenAI, and it never marks up or resells model usage.

How is Codex in AQ different from OpenAI's cloud Codex tasks?

Three ways, as of July 2026: execution runs on infrastructure you control rather than OpenAI's sandbox, you get a live interactive terminal rather than a delegated task that returns a result, and the session is multiplayer, so teammates can watch and steer it. The trade is that you bring the infrastructure (or use an AQ-managed VM).

Can I run Codex and Claude Code side by side?

Yes, that is one of the most common AQ patterns: two workspaces, two worktrees, same repository, one running Codex and one running Claude Code, sometimes on the very same task to compare results before choosing which PR to take.

Does Codex cost extra inside AQ?

Your model usage is billed by OpenAI exactly as it would be locally. AQ itself has a Free plan (bring your own VM) and a Team plan at $100 per user per month promotional (standard $200) that includes a dedicated AQ-managed, always-on VM.

Can teammates on Windows or Linux join a Codex session?

Yes. AQ is browser-based, so the same live workspace (terminal, editor, preview) works from macOS, Windows, Linux, and mobile browsers. The CLI executes on the workspace VM, not on anyone's device.