Nobody makes an account by just signing up. The first account in your own Agentic Cloud comes through the one-time link its installer prints; everyone after that comes through the workplace’s invite link, and an admin approves each person. The link opens the sign-up page: type your name, choose an address and a password, and press Create account. You are signed in at once. Nothing is emailed to you and there is no link to wait for.

The address is a new one, on your Agentic Cloud’s own domain. You type only the part before the @, and the page tells you as you type whether the name is free. Choose it with care: it is the name you sign in with, and once an admin has set up email it is a real inbox that people can write to (see Email).

The first account’s workplace is created with it and named after its owner for now; everyone invited joins that one. You can rename it whenever you like.

Your Chief of Staff is your own agent across the whole workplace, and Home is its conversation. The first time, it is empty but for one line: tell it what you are working on, and it puts a team on it. Type your answer as you would brief a person: the outcome, and anything the team should know.

A new workplace is empty until this point. There are no agents yet; the Chief of Staff brings them in for the work you describe.

Your agents think with a subscription you already pay for, so the Chief of Staff’s answer to your first message asks for it. Under its answer is a card with two rows: Claude for a Claude Pro or Max subscription, Codex for a ChatGPT Plus or Pro subscription. Tap the one you have. You can add the other kind later, an OpenCode Go subscription too, and more of each. It is the only thing to connect: your agents’ computers already run in your company’s cloud account.

Claude: press Open the Claude sign-in, approve on Claude’s own page, and paste the code it gives you into the field. The code is two long strings joined by a #, and you paste it whole.

ChatGPT: copy the code shown on the screen, press Open the ChatGPT sign-in, enter the code on OpenAI’s page and approve. The screen notices by itself and takes you back to the conversation.

Either way you sign in on the provider’s own page, and your password never touches the platform. The subscription is yours alone: no teammate or admin can see or use it. Anything you write before signing in is kept, and the Chief of Staff starts on all of it once the plan is in. What a plan can run and how usage is shown is in Your subscriptions.

When the plan lands, the conversation says so, then Getting your Chief of Staff ready… while its own computer is built. That happens once and takes between fifteen seconds and a minute. Then it thinks about what you wrote.

It brings in only the agents the work needs, usually two, on the plan you signed in. Each step appears as a line in the conversation, and then its reply: the plan, with one line per agent and what each one is doing. Under the conversation, the Arriving card shows each agent as its computer is built, which takes a minute or two. The agents are named for you, and each joins the workplace’s one list (see One list of agents).

You do not have to watch. The first result arrives in the same conversation: the agent’s own answer, then the Chief of Staff’s line with what is done, a link to the page if there is one, a line for each other agent’s result, and what it can do next. Answer it there, or tap the faces beside the message box and any agent to talk to it yourself. From then on everything in this guide applies.

Each agent works on its own computer. Close the tab, and when you come back the faces beside the message box show which agents have replied. Work that touches protected code comes back as a proposal for you to approve, unless you have turned on auto-merge for your agents. Read The proposal loop when your first one arrives.

Home stays the Chief of Staff’s conversation. Ask it to bring an agent in, hand out work or read your mail, and it reports back here.

Tap its face or its name at the top to open the Workplace in a window over the conversation: the people here and their agents, your messages and mail, your subscriptions, your auto-merge switch, and a row each for notes, Drive, code, knowledge, sites and keys.

The faces beside the message box open your agents in one list, with Waiting for your decision above them once an agent has proposed something. To catch up after a day away, ask the Chief of Staff. Your picture at the top right opens a menu: Settings, the people here, and every app launched here.

Open any Workspace address and sign in with the name you chose and your password. The bare name works; so does the whole address. With authenticator security on, you are asked for the current code next.

Forgot your password? sends a reset link to your recovery email. A new account has none until you add one, and without it a forgotten password means a lost account: nobody can reset it for you. Add one today, in Settings. The reset link and the recovery email both need email: until an admin has set it up under Mail, Settings says so.

Settings, from your picture on Home, holds your name and picture, your address, and the Access card. Three things there are worth doing on day one.

• Add a recovery email. It is the only way back in after a forgotten password. You confirm it through a link that lasts thirty minutes.
• Turn on authenticator security if you want a second step at sign-in. Save the recovery codes it shows you; each one replaces the authenticator once, and they are shown only there.
• Check your sessions. Every signed-in browser is listed, and you can sign any of them out.

The same screen sets which agent, if any, is woken when mail arrives (see Your address).