Steer a run as a team

1. Share the invite link

Open Settings → Organization and press Copy invite link. The link carries your org's invite code. Anyone who opens it lands on Join a workspace on Rush Cloud, signs in with Phoenix ID if they have not, and presses Join workspace.

Admins also see an email field with a Member or Admin role and an Invite button.

2. Know who can join

An org is tied to its email domain. A teammate joins when their email domain matches the org's and their code is valid:

Any member can share the code. An admin can regenerate it if it leaks, and can change a member's role or remove them from the same page.

3. Open a Project together

There is no separate Project member list: everyone in the org can see every Project, start runs in it and add automations. A Project's Settings → Members says so and links to the org's members.

The Project's runs, and Runs in the sidebar, show each run's status and started by with the person's name. A run's page lists Started by in its details, and the transcript puts the starter's name on each message they sent.

4. Follow it live

Open a teammate's running run and its transcript streams as it happens, the same view its starter has: what the agent says, each command, each edit. Any member of the org can read any run in it.

5. Queue the next instruction

You do not have to wait for the agent to finish. Type in Follow up on this run and send it. If a turn is running, the console answers Message queued. It will be delivered when the current turn ends. and the message appears under Queued messages, marked Queued, then Delivering and Delivered.

• A queued message can be edited or removed until it is delivered; a failed one can be retried.
• A run holds at most 20 queued messages. One more is refused with QUEUE_FULL.
• Files cannot wait in line. A message with files waits until the turn ends.
• Stop ends the current turn early, so the next queued message goes through now.

Sending, editing and removing follow-ups is open to the run's starter and to org owners and admins.

Queuing is how the console and the CLI behave. A follow-up sent with an API key while a turn runs is refused with 409 TURN_ACTIVE instead; see Runs and sessions.

6. Watch the live display

The Live display card on the run page shows the run's own computer and browser, read-only, with Live · read-only above it. The run's starter and org owners and admins can open it.

7. Take control

When the agent is stuck on a login, a CAPTCHA or a page it cannot read, press Take control. You now drive the browser yourself; the card shows You are controlling with a countdown.

• Control is a lease of 5 minutes. The console renews it every minute while your tab is visible, so it lasts as long as you are working and ends soon after you leave.
• One person holds it at a time. A teammate who tries meanwhile sees Held by with your name and when the lease ends; the API answers them 409 CONTROL_HELD with the holder and expires_at.
• Taking control is open to the run's starter and to org owners and admins.
• Release control hands the browser back to the agent. The run's starter or an admin can also release a lease someone else holds.

Then send a follow-up saying what you did, such as "I signed in; continue from the dashboard."

Watch and steer covers the live display and follow-ups in depth, and Isolation explains what separates one org's runs from another's.