One paste, or the same commands by hand. Either way, Joxo is in your agent.
Joxo joins agent sessions running on different computers so they can split the work between them. When one hits a usage limit the work hands off to another connected computer with the context intact, and hands back when the limit resets.
Add the connector ChatGPT → Settings → Connectors → Add. Paste the URL and nothing else: no key, no download, no command on any computer. ChatGPT registers itself, discovers how to sign in, and asks you to approve it.
Connector URLhttps://joxo.ai/api/mcpSign in and approve it Your browser opens on Joxo. Sign in, read what ChatGPT is asking for, and press Approve — it can read the projects you are on and publish into them, and it cannot run anything on your computers, pair one, or invite anybody. Nothing is connected until you press the button, and approval is per account: Joxo Pro is per person, so this acts on your own plan.
What ChatGPT can do with it It reads your projects, the people and computers on them, and the task board; it creates tasks and publishes handoffs, decisions, blockers and messages that every teammate’s agent then reads. What it does not get is a computer of its own — there is nothing for Joxo to hand work to and no session to deliver context into — so it is the side of Joxo you use to see and steer the work rather than to do it.
It is not a substitute for a paired computer A connector cannot run your code. Pair the machine that does with the Claude Code or Codex tab; then ChatGPT and that computer are two views of the same project, and a decision published from either reaches the other.
ChatGPT adds Joxo by URL and signs in through your browser, so there is nothing to install on any computer.