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

# Conversation

> Ask in plain language, follow the work as it happens, and step in when needed.

The **Conversation** view is where you work with agents. Describe what you want done; the agent plans the steps, calls the tools and apps it needs, and reports progress as it goes.

## Asking for work

Describe the outcome, not the clicks. Good requests say:

* **What** you want produced — a report, a draft, a page, a list of leads.
* **Where** it should go — an email, a Slack channel, a page in this project.
* **When** — right now, or on a schedule (the agent can set up a [routine](/docs/features/routines.md) for you).

If something is ambiguous, the agent asks you a question in the conversation before it continues.

## Files

Attach files to a message to give agents something to work from — spreadsheets, documents, images, and more. Each file can be up to 200 MB. Files agents create show up in the conversation as cards you can preview and download.

## Choosing a model

Use the model picker under the message box to choose between **Lite**, **Standard**, **Smart** (default), and **Ultra**. Lighter models are faster and use fewer credits; Ultra is the most capable and is available on paid plans. See [Plans and credits](/docs/plans-and-credits.md#choosing-a-model).

## Approvals in the conversation

When an agent prepares something that acts on your behalf — an email, a calendar invite, a social post — it appears as an [approval card](/docs/features/approvals.md). Nothing is sent until you approve it.


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