Cloud and subscriptions

The platform runs as your own Agentic Cloud, in your company’s own cloud account (see Your own Agentic Cloud). Each cost is billed by whoever provides it:

• Us: nothing. Your own Agentic Cloud is free for now: no licence, no key and no charge from us.
• Cloud: your cloud provider bills your own account directly for the agents’ machines, your sites, your files and the Workspace itself.
• AI: your Claude, ChatGPT or OpenCode subscription, billed by Anthropic, OpenAI or OpenCode as it always was.
• Voice-to-text: your workplace’s own OpenRouter key, billed by OpenRouter for each minute of speech (see How to turn on voice-to-text).
• Anything else you connect, such as a GitHub account: on that provider’s own terms.

Three kinds of thing run in your cloud account for your workplace.

• One machine per agent. It keeps its files between turns, sleeps when idle and wakes in seconds when there is work. See The machine and Sleep and wake.
• The sites your agents deploy, each on its own small machine. See Production sites.
• Your files: the ones sent in chats, put on notes, proposals and knowledge pages, and everything in Drive. See Where files live.

Repositories live on your Agentic Cloud’s own code host, in the same account, unless you use a host of your own; see Where code lives and the Privacy Policy for the full split.

Admins see how it is all doing under This copy in Workplace settings: one line each for the Workspace, its backups, the disk, the code host, previews and sites, and mail coming in.

Several agents can share one seat, and more seats run more agents at once (see Seats).

A subscription you have signed in is called a seat: a Claude Pro or Max plan, which runs agents on Claude Code, a ChatGPT Plus or Pro plan, which runs them on Codex, or an OpenCode subscription, added with its key, which runs them on OpenCode (see Sign in a subscription). The agents run the official Claude Code, Codex and OpenCode tools on their own machines. The platform gives them the machines, and the plan itself stays between you and its provider. It is yours alone. It runs only your own agents, and no teammate, admin included, can see its credential or its usage. Several of your agents share one seat, and you can use the same seat in several of your workplaces at once.

Your subscriptions appear as a strip in Workplace on Home, one tile per subscription with its usage, and Add sub opens Your subscriptions. There you sign in another, or bring in one you use elsewhere with Use here. With more than one seat of a provider, Make primary here puts one first; when it is paused, your agents fall back to the next ready seat of the same provider. Unassign from this workplace keeps a seat working in your other workplaces, while Remove subscription everywhere takes it out of all of them and deletes the platform’s copy of its sign-in. It does not sign you out at Anthropic or OpenAI.

A subscription has windows: an amount for a short period and an amount for the week. A Claude seat shows 5 hours, 7 days and a weekly window per model. Each shows the percentage used, when it resets, and a mark where an even pace would be by now. A high percentage late in the window is fine; a high one early means the agents will hit the limit before it resets, and the pace line forecasts when. An agent that reaches the limit pauses at Session limit and picks up again on its own once the window resets. When Claude pauses a subscription, usually a billing matter on their side, the seat’s screen says so and offers Open Claude billing. OpenCode keeps its usage in its own console, so an OpenCode seat shows none here, and your agent says when a limit is reached.

Voice-to-text lets anyone in the workplace speak a message and check the words before sending. Your workplace pays for it with its own OpenRouter key, so it stays off until an admin saves one. Until then the microphone says it is not set up yet.

An admin makes a key at openrouter.ai and adds credit there. On the Workplace screen, under Voice-to-text, they press Turn on with an OpenRouter key and paste it. OpenRouter checks the key before it is kept, and no agent ever sees it.

The same section picks the model that listens. Each one shows its price a minute, and GPT-4o mini Transcribe is the default. Only an admin, in the Workspace, can change the key or the model.

When the account runs out of credit, voice-to-text pauses for everyone and the microphone says why. Add credit at openrouter.ai and it comes back by itself within a couple of minutes.

Every agent’s machine has a size: how many processor cores, whether they are shared or dedicated, and how much memory. The standard for a new agent is two dedicated cores and 4 GB of memory, which is a good size for almost all work. The ladder runs from one shared core with 2 GB up to eight dedicated cores with 16 GB.

• Shared cores: one, two or four cores with 2, 4 or 8 GB. A shared core is throttled hard under sustained load, and the same job can take many times longer there.
• Dedicated cores: one, two, four or eight cores with 2, 4, 8 or 16 GB. The standard is in this list.

Memory is a wall, not a slope. There is no swap, so work that outgrows the memory is stopped rather than slowed, and the fix is one step up the ladder.

An agent’s owner, whatever their role, resizes that one agent from its own screen, in its Machine row. A resize is refused while the agent is mid-turn. An awake machine restarts and stops for about a minute; an asleep one wakes up on the new size. Files are kept either way.