Step 10

Organizations and the organization's AI

Organizations

An organization is the company that owns the work brains. It is in the left menu under Settings → the organization's name and has the parts Overview, Members and Domains (domains for admins only).

  • Brains of the organization: the list of brains with their owner. When someone leaves or changes teams, an admin uses Hand over to give the brain to a successor, or archives it.
  • Chat in the app: an admin turns the chat on or off for the whole organization. Below it is Internet in the chat (only where the installation has it): On, except in brains with protected data, On everywhere, also in brains with protected data, or Off. Only the search query and the page address go to the web, the content of the brain does not.
  • People in the organization: an e-mail, a role (Member or Admin) and Add. A person becomes a member after accepting the invitation. Anyone who is only in a brain of the organization is listed as a Brain member: no rights in the organization and no access to its other brains.
  • Company domains: once a domain is verified (a TXT record in DNS), people with a company e-mail can join the organization themselves, or ask to join.

A new brain gets its organization in the wizard already: in the Basics step, the Organization field offers Personal, your organizations and Add a new organization. The same holds when uploading your own brain.

The organization's AI

Organization AI (in the left menu under the organization) is the company's own AI: the local AI on its graphics card, a runner with Claude or its own key. Only the organization's owner and admins see it, and only where the plan allows it. Tabs:

TabWhat is there
Runnersconnected runners and keys with their state (Online, Offline, Paused …), Add AI, Pause, Finish and stop, Rotate key, Revoke
Routingthe setting: which feature should run on which runner and what happens when the runner is not running. Under the table the overview Where the features run right now
Datathe least sensitivity of every brain and where the runners' data goes, see AI and data
Jobsthe job queue and the history
Usagethe plan's limits, and requests and tokens by day and by feature
Logstechnical events of the runners for 30 days, without content

Add AI

  1. Runners → Add AI, pick the Kind of AI:
    • Own server with local AI (the default): a runner on a server with a graphics card,
    • Runner with Claude: a runner that works with Claude on the company's server,
    • Own Anthropic key: the chat and other features on the company's key,
    • Server with the OpenAI interface: advanced, a server given by its address (vLLM, Ollama).
  2. For a runner, click Create command and run the command on the server where the runner is installed. The command works once and only for a short time. The page sees by itself that the runner has connected (The runner is connected). How to install the runner is on the Runner guide page.
  3. For a key or a server, fill in the key or the address, Test and Add.

The runner connects only from the server outwards, nothing is opened into the company.

Routing

The Routing table lists every feature (chat, change summary, document descriptions, meeting transcription and summary, AI process draft, agents, images …) and where it runs. The default is the onlybrain account. The When the runner is not running column offers:

  • onlybrain account with a cap: the fallback in the onlybrain cloud takes over, up to a cap. The cap is a percentage of the default limit, 0 turns the fallback off.
  • Wait for the runner: the job waits.
  • Fail: the job ends with an error.

Data labelled company's own AI only or local AI only never goes to the fallback in the cloud. Unsaved changes are marked unsaved, Save routing saves them, Discard changes drops them.

Under the table is the overview Where the features run right now. Nothing is set there: it shows the result of the routing at this moment, where each feature really runs and why (runs, fallback, queued, waits, not allowed, off). Besides the routing it depends on the local AI's mode and calendar and on the data labels.

Jobs and history

Jobs has two views: Queue (what waits and runs) and History (finished jobs and chats of the last 30 days). The history shows only facts about the work: who, in which brain, when, on which AI, for a chat the number of messages. Neither the content nor the title of a conversation is there. The User filter shows the work of one person.

Usage by plan

Usage shows the Plan limits on the onlybrain account: what percentage of the daily or monthly chat limit is used and when it renews. All members of the organization share the limit, and only the chat on the onlybrain account counts, work on the organization's own AI does not. Once the limit is used up, the chat on the onlybrain account waits for the renewal, the own AI keeps running. Higher limits: see the plans leads to the pricing.

Below the limits are requests, tokens and fallback requests for 7 or 30 days, By day and By feature.

The local AI on a graphics card has a page of its own, see Local AI.