> ## Documentation Index
> Fetch the complete documentation index at: https://learn.oxom.de/llms.txt
> Use this file to discover all available pages before exploring further.

# Chatting with the Agent

> Ask questions, attach files, address connections, have answers read aloud.

Open **Agent** in the sidebar. If the chat is empty, you see three suggestions to try.

## Writing

* **Enter** sends the message, **Shift+Enter** starts a new line.
* Write the way you talk. You don't have to learn commands.
* If something is unclear, the Agent asks one targeted question instead of guessing.
* It replies in your language: German, English or Dutch.

## Attaching files

With the paper clip **Datei anhängen** (attach file) you give it images (JPG, PNG, WebP, GIF), PDFs or calendar files (ICS), up to 20 MB.

* **Images** it looks at, e.g. a thumbnail it should suggest titles for.
* **PDFs** it reads, e.g. a briefing or a contract.
* **ICS files** it reads as appointments.

You don't have to attach files from your HUB. Just say which one you mean: “Read the briefing in the Voltaro folder.”

## Viewing the steps

While the Agent works, each step appears below its reply, e.g. “Kurzlink läuft …” (short link running). One click expands what it entered. That way you see exactly what happened. **Stoppen** (stop) cancels.

## Using answers

When you hover over a reply:

* **Kopieren** (copy) puts it on the clipboard.
* **Vorlesen** (read aloud) reads it to you, up to 5,000 characters.

Voice input is not available in the chat yet.

## Clearing the chat and continuing

* **Chat leeren** (clear chat, the trash icon at the top) starts a new conversation. What it has [remembered](https://learn.oxom.de/en/werkzeuge/agent/gedaechtnis.md) stays.
* Your chat is the same on all your devices. If you continued writing elsewhere, a notice with **Fortsetzen** (continue) appears.

## Addressing connections with @

When you type `@`, the chat suggests connections, e.g. `@guide` for these guides. **Tab** inserts the suggestion. This only works if the connection is set up for your workspace. Otherwise the Agent carries on as usual.

## Limits

* Per message the Agent takes up to 8 steps. Larger tasks are best split up.
* Each person can send 30 messages per minute.
* If it declines something, it says „Dabei kann ich nicht helfen.“ (I can't help with that).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.