| Administrator | A person who manages the whole of AAS. It is one of: a person who received the administrator role on the sign-in server, a person designated on the "Users & permissions" screen, or a person the operator designated at installation. |
| Allowance | The act of an administrator permitting each AI agent workspace to receive packages from the "Restricted" namespace. It is a record inside AAS and has nothing to do with billing. |
| Approval settings | Settings that decide how approval is obtained before a risky command runs. Only administrators change them. |
| Auto upgrade | A function that automatically upgrades profiles that have not been modified to the new version of a built-in skill. |
| Automatic approval (no admin review) | A way of publishing the publish request for a skill made by the owner without administrator approval. |
| Backup copy | A copy of the content from before a change, kept for a set period. |
| Band | A one-line message shown at the very top of the screen. Notices, restart guidance, and rollout progress steps appear as bands. |
| Canary | A test AI agent workspace that is switched to the new version first. |
| Change log | A screen that records who changed what and when. Values are not recorded. |
| Channel | One messenger connected to a profile. You work with it on the "Channels" tab of the profile detail window. |
| Collection | The act of reading and saving the dashboard numbers at regular intervals. |
| Container | A way of running a program bundled together with the files it needs. |
| Dashboard | The first screen, which gathers the status of AI agent workspaces and the items that need action on one screen. |
| Data volume | The disk that an AI agent workspace uses. It remains even if the workspace is deleted. |
| Department | An organizational unit that a user belongs to. It becomes the audience for department notices and the scope of a team manager. |
| Dictation | A function that turns speech into text. You use it with "Voice input" in the message box. |
| Drawer | The detail pane that opens on the right side of the screen. This manual calls it this. It opens when you click a row in a list. |
| Export file | A file that keeps the settings, scheduled jobs, skills, and memory of a deleted AI agent workspace. |
| Fallback | The model that is used instead when the main model cannot answer. |
| False positive | A decision in which a guardrail wrongly blocks normal text. |
| Force disconnect | An administrator disconnects another person's Claude Code connection and writes a reason. |
| Gateway | The channel through which an AI agent receives messenger messages and scheduled jobs. Only administrators turn it on and off. It is used together with "Messenger connections" on the screen. |
| General user | A person who is signed in and is not an administrator, and is neither the owner nor a team manager of the AI agent workspace in question. |
| Group | A unit that collects similar items. There are the "Kind" groups of the change log and the screen groups of the left menu. |
| Guardrail | A safeguard that checks a question before it goes to an AI agent and blocks it if it goes against policy. |
| Guardrail category | A classification of the reasons a guardrail blocked something. Example: jailbreak attempt. |
| Image | A set of files used to create the container that runs the server or Hermes. |
| Inbox | The place where approval requests, user requests, and notices gather. |
| Internal certificate (CA) | A certificate made by a Certificate Authority that the company runs itself. A PC must trust this certificate to connect to internal HTTPS addresses. |
| Internal wiki connection | A connection that lets an AI agent read and write pages in an internal wiki such as Docmost or Outline. |
| Kanban | A board inside the Office that divides work into cards and manages them. |
| Kubernetes | A platform that runs and manages server programs across several computers. It is the environment in which AAS is installed. |
| Kubernetes namespace | A division that separates resources inside Kubernetes. The resource limits of an AI agent workspace are set in this unit. |
| Main model | The AI model a profile uses to make answers. |
| Main profile (main) | The profile a user receives first. Its name is main. |
| Manual token | A Claude Code connection token used where there is no browser. Choose the expiry from 7, 30, 60, or 90 days. |
| Mark as seen | Marking an item in "To check" as read so that it is removed from the list. It does not fix the problem, it only closes the notice. |
| Messenger connection | A way to talk to an AI agent through a messenger such as Telegram or Slack. The program that accepts the connection is the gateway. |
| Mirror | A copy of a publicly released skill. It is in the community area. |
| Namespace | A division that separates packages inside the Skill Hub. There are shared, restricted, and community. It is different from a Kubernetes namespace. |
| New version | The state in which a higher version has been published to the Skill Hub and can be upgraded to. |
| Node | The server computer on which AI agent workspaces run. Several workspaces run on one server. |
| Notice | A message that announces maintenance, outages, and news of new features in a band at the top of user screens for a set period. |
| Office | A 3D screen where several AI agents and people work together in one room. |
| Operator | The person who installs AAS and manages the server. They may or may not be the same person as the administrator. |
| Owner | The user who has that AI agent workspace. The same as workspace owner. |
| Ownership transfer | Changing the owner of an AI agent workspace to another person. |
| Package | The unit of distribution published to the Skill Hub. |
| Personality (SOUL) | The text that decides what kind of AI agent a profile is. |
| Profile | The set of settings for one AI agent inside an AI agent workspace. Each workspace starts with one main profile (main). |
| Publish request | A request in which a user asks to put a skill they made into the Skill Hub. |
| Read aloud | A function that plays an AI answer as speech. |
| Restricted | A Skill Hub namespace that an administrator must allow for each AI agent workspace before it can receive its packages. |
| Role template | A template for the personality (SOUL) of each company role. Administrators make them and use them for new profiles. |
| Rollout | Switching users' Hermes to a new version, one after another. |
| Scheduled job (cron) | Work that an AI agent does at a set time. |
| Secret | A key or password put into a profile. The value is not shown again on the screen. |
| Session (chat session) | One set of conversation with an AI agent. The "Usage" screen counts them. |
| Session (sign-in session) | The signed-in state that continues after you sign in. It ends if nothing is done for a set time. |
| Shared | A Skill Hub namespace that every AI agent workspace can receive. |
| Skill | A bundle that teaches an AI agent how to do a particular job. |
| Skill Hub (SkillHub) | The place that stores skill packages and lets you install them into profiles. |
| Skill map | A screen that shows as a diagram where a skill is called from and what it calls. |
| Team manager | A person an administrator designates for each department. They treat the AI agent workspaces of people in the same department as if they were the owner. |
| To check | The list of items that need action, shown on the dashboard and in profiles. |
| Token (API key or token, manual token) | A secret key used to connect. It is different from a token, the unit of text that a model reads and writes. |
| Token (usage) | The unit of text that a model reads and writes. Input tokens are the text and instructions the user sent. Output tokens are the answer the AI wrote. |
| Toolset | A unit that groups the tools an AI agent can use so you can turn them on and off together. |
| Tooltip | A small explanation that appears when you hover the mouse over something. |
| Usage | A screen that shows chat sessions and token usage. |
| Work instructions (AGENTS) | The text that describes how a profile works. |
| Workspace owner | The user who has that AI agent workspace. The same as owner. |