# Settings

{/* AUTO-SYNCED SOURCE: this page lives in apps/app/src/modules/agent/docs/ and is mirrored into the docs app by `bun sync:module-docs`. Edit it in the module, not in apps/docs. */}

<Lead>
The assistant has two settings levels: the workspace settings under **Settings > Assistant** (Einstellungen > Assistent) define what the assistant may do for the whole team and how it presents itself (owners and admins only). The personal settings under **Account > Assistant** (Konto > Assistent) belong to each person individually.
</Lead>

## Workspace settings

The **Settings > Assistant** page (`/settings/agent`) is for owners and admins and is organized into these sections:

<DefinitionList>
  <DefItem term="Header card">At the very top: your assistant's name, which modes are running, and whether it acts on its own. **Umbenennen** (rename) gives it a custom name matching your office. It then appears throughout the assistant area. Without a custom name it is simply called "Assistent".</DefItem>
  <DefItem term="Modi (modes)">One card per mode with its cost ratio and the model behind it. You enable or disable Standard and Premium here, Lite only the owner. One mode always stays active: the last one cannot be switched off. Switching a mode off asks first, and running chats keep their mode. Details: [Modes and costs](/tools/agent/modi-und-kosten).</DefItem>
  <DefItem term="Von selbst tätig werden (acting on its own)">Workspace-wide switches for the morning briefing and the proactive suggestion cards on the dashboard. What is off here cannot be enabled by any individual person.</DefItem>
  <DefItem term="Gedächtnis langer Chats (memory of long chats)">When a long-running chat compresses its oldest messages into a summary: **Früh** (early, saves the most), **Ausgewogen** (balanced, the default), or **Spät** (late, the history stays verbatim the longest and costs more in long chats). The visible history always stays complete, you can scroll back at any time. Only what the assistant re-reads with every new question gets summarised.</DefItem>
  <DefItem term="KI-Modelle (AI models)">Pin a fixed model per mode, separately for text and image. Without a selection, REOS AI decides automatically. Bringing your own provider keys (BYOK) is listed but still locked.</DefItem>
  <DefItem term="Weiterführend (further reading)">Two slim links at the end of the page: [Plugins](/tools/agent/plugins) (where you enable or lock every skill individually) and the [knowledge vault](/tools/agent/wissen-und-gedaechtnis). Both are areas of their own with their own entry in the assistant sidebar; this page only holds the way there.</DefItem>
</DefinitionList>

Since 5 September 2026, plugins and the knowledge vault sit at the end of the page instead of the top: both are links, not settings, and they stood in front of what this page actually decides. The name moved into the header card instead.

## Plugins instead of the permissions console

Until 5 September 2026 the **Berechtigungen verwalten** (manage permissions) button opened a separate console. It has merged into the plugins: every group of the console is now a workspace plugin under **Assistant > Plugins**, every permission in it a skill. Your previous permissions were carried over, and the old address redirects there. Structure, rights, and the approval of the personal onOffice AI Connector: [Plugins](/tools/agent/plugins).

The principle remains: **What is locked, the assistant does not execute.** If someone asks for a locked skill, the assistant honestly says it is not enabled (see [locked skills](/tools/agent/faehigkeiten#locked-skills)).

<Info title="Confirmations still apply">
Enabled does not mean unasked: even with full permissions, each person confirms change actions individually in the chat via the confirmation cards.
</Info>

## Personal settings

Under **Account > Assistant** (`/account/agent`) each person configures their own assistant:

<DefinitionList>
  <DefItem term="Plugins">A jump to your personal plugins under Assistant > Plugins. Since September 2026 the personal **onOffice AI Connector** is set up and disconnected there, no longer here. Details: [Plugins](/tools/agent/plugins).</DefItem>
  <DefItem term="Gedächtnis (memory)">Your personal memory entries: view, add, delete individually, or forget everything. Details: [Knowledge and memory](/tools/agent/wissen-und-gedaechtnis).</DefItem>
  <DefItem term="Verhalten (behavior)">Your day-to-day toggles: floating orb on or off, the **Dazu fragen** (ask about this) chip on or off, morning briefing and proactive cards on or off.</DefItem>
</DefinitionList>

<Warning title="Workspace switches take precedence">
The personal toggles for briefing and proactive cards only take effect if the workspace allows these features under **Proaktivität** (proactivity) in the first place. If a feature is off workspace-wide, it stays off for everyone.
</Warning>

## Who may change what?

The workspace settings under **Settings > Assistant** can only be changed by owners and admins. The personal settings under **Account > Assistant** are open to every person for themselves. Who may do what with the plugins is on the [Plugins](/tools/agent/plugins) page.

## Further reading

What the individual skills do is covered in the [capabilities reference](/tools/agent/faehigkeiten), who enables them in [Plugins](/tools/agent/plugins). How mode decisions affect costs is explained in [Modes and costs](/tools/agent/modi-und-kosten). Back to the [overview](/tools/agent).