A.1. Permission Table
Overview
This table shows at a glance who can do each AAS function. It was made by checking the menus on the screen and the permission checks on the server. Use it when a user asks about "a button that is missing" or "a 403 error", and when you decide who to give a permission to.
Permission Roles
| Role | Who it is |
|---|---|
| Administrator | A person who received the administrator role on the sign-in server (SSO), a person designated as an administrator on the "Users & permissions" screen, or a person the operator designated as an administrator at installation |
| Workspace owner | The owner of that AI agent workspace |
| Team manager | A person an administrator designated as the team manager of a department. They treat the AI agent workspaces owned by people in the same department (team member AI agent workspaces) as if they were the owner. |
| 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 |
One person can belong to several roles. For example, an administrator is also the owner in their own AI agent workspace, and a team manager is also the owner in their own workspace.
In every table in this chapter, the "General user" column means only the definition above. The functions in the "Accounts and Organization" table do not depend on a workspace, so owners and team managers have the same values as general users. "Not allowed" means the function cannot be used. It is not shown, or it is refused when clicked.
Warning: In an installation where sign-in is turned off, everyone is treated as an administrator. This installation is for testing and can be reached only from that computer.
Warning: What is done through a Claude Code connection (token) cannot use administrator permissions, even for an administrator's connection. Do administration tasks by signing in through the browser.
Note: Hiding menus and buttons on the screen is a convenience. The server decides permissions. Even if someone goes directly to the address, a non-administrator sees "Only admins can use this screen."
Accounts and Organization
| Function | General user | Administrator |
|---|---|---|
| The "Users & permissions" screen; adding, editing, disabling, and deleting users | Not allowed | Allowed |
| Managing departments, designating team managers | Not allowed | Allowed |
| Forcing a user to sign out | Not allowed | Allowed |
| Company name and the AI agent introduction text | Not allowed | Allowed |
| Slogan and goal of your own department ("Settings › Company info") | Allowed | Allowed |
| Creating and editing role templates | Not allowed | Allowed |
| Posting, taking down, and extending notices | Not allowed | Allowed |
| Handling user requests in the inbox ("Mark handled", "Decline") | Not allowed | Allowed |
| Dashboard of all AI agent workspaces | Not allowed | Allowed |
| The full change log | Not allowed | Allowed |
| Creating your own AI agent workspace | Allowed within the limit | Allowed |
| Creating another person's AI agent workspace, changing the count limit, designating a new owner | Not allowed | Allowed |
| Rolling out the new Hermes image to users | Not allowed | Allowed |
| "Test" for a model connection | Not allowed | Allowed |
| Recording status, owner, and notes on the API keys & tokens screen | Not allowed | Allowed |
| Creating and running Kanban boards, renaming and archiving boards | Not allowed | Allowed |
| Viewing and force-disconnecting everyone's Claude Code connections | Not allowed | Allowed |
| Creating and disconnecting your own Claude Code connections | Allowed (in installations with sign-in on) | Allowed |
Functions Inside an AI Agent Workspace
Read this table based on the "AI agent workspace you are viewing".
| Function | General user | Workspace owner | Team manager | Administrator |
|---|---|---|---|---|
| Deleting an AI agent workspace, restarting a stopped AI agent workspace | Not allowed | Allowed | Not allowed | Allowed |
| Editing the personality (SOUL) and work instructions (AGENTS), applying a role template to the personality (on the "Personality (SOUL)" tab) | Not allowed | Allowed | Allowed | Allowed |
| Turning on and off, and renaming, MCP servers, toolsets, and skills | Not allowed | Allowed | Allowed | Allowed |
| Editing profile settings files | Not allowed | Allowed (except some items, see below the table) | Allowed (except some items, see below the table) | Allowed |
| Entering and deleting secrets, saving messenger connections and pairing | Not allowed | Allowed | Allowed | Allowed |
| Creating a profile | Not allowed | Allowed | Allowed | Allowed |
| Deleting a profile | Not allowed | Allowed (except the main profile) | Allowed (except the main profile) | Allowed |
| Starting, stopping, and restarting the gateway (the "Gateway" section of the profile "Overview" tab) | Not allowed | Not allowed | Not allowed | Allowed |
| Changing the main model | Not allowed | Sends a request only | Sends a request only | Allowed |
| Sending a request to an administrator (model change, profile limit, and so on) | Not allowed | Allowed | Allowed | Not applicable (handles it directly) |
| Creating, editing, pausing, and immediately running scheduled jobs | Not allowed | Allowed | Allowed | Allowed |
| Editing skill documents, creating new skills | Not allowed | Allowed | Allowed | Allowed |
| Viewing the usage screen | Allowed in AI agent workspaces they have access permission for | Allowed | Allowed | Allowed |
| Viewing the change log of this AI agent workspace | Not allowed | Allowed (records from after the current owner became the owner) | Allowed (records from after the current owner became the owner) | Allowed (all records) |
Note: In the settings file, only administrators change the items for model, provider, auxiliary model, gateway, terminal execution location, channel authorization, and command approval. Owners and team managers can edit all other items. If you try to change a blocked item, the screen tells you which item it is. Owners and team managers also cannot edit the hidden root profile (default). This restriction does not apply to administrators.
For how to start and stop a gateway, see step 3 of 7.1. Operations Checks. For handling model change requests, see 5.3. Handling User Requests.
Skill Hub
| Function | General user | Workspace owner | Team manager | Administrator |
|---|---|---|---|---|
| Viewing the Skill Hub list | Allowed | Allowed | Allowed | Allowed |
| Installing and deleting skills, restoring archived skills | Not allowed | Allowed | Allowed | Allowed |
| Viewing "New versions" and upgrading to a new version | Not allowed | Allowed | Allowed | Allowed |
| Allowing and removing the allowance for the "Restricted" namespace | Not allowed | Not allowed | Not allowed | Allowed |
| Creating a publish request | Not allowed | Allowed | Allowed | Allowed |
| Viewing the publish request list | Not allowed | Allowed (those of their own workspace) | Allowed (those of that workspace) | Allowed (all) |
| Approving and rejecting publish requests | Not allowed | Not allowed | Not allowed | Allowed |
| Setting automatic approval of owners' skills | Not allowed | Not allowed | Not allowed | Allowed |
| Setting auto upgrade | Not allowed | Not allowed | Not allowed | Allowed |
| Publishing a zip directly, hiding from the hub and showing again | Not allowed | Not allowed | Not allowed | Allowed |
| Reclassifying skill tags, retranslating skill descriptions | Not allowed | Not allowed | Not allowed | Allowed |
Note: Automatic approval of owners' skills applies only to requests made by owners. An administrator approves requests made by an administrator directly. For details, see 6.2. Reviewing Publish Requests.
Menu Visibility
| Menu | General user | Workspace owner | Team manager | Administrator |
|---|---|---|---|---|
| Admin group (Users & roles, Role templates, Notices) | Not shown | Not shown | Not shown | Shown |
| Records › Change history | Not shown | Shown (title: My AI Agent Workspace change history) | Shown (title: Team AI Agent Workspace change log) | Shown |
| Settings › Company info | Shown (only their own department's text) | Shown (only their own department's text) | Shown (only their own department's text) | Shown (all) |
The other menus (Dashboard, Chat, My AI Agents, Skills, Automation, Records › Usage, Settings) are shown to every signed-in user. What can be done inside them follows the tables above.