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

# Chatting with the AI

> Direct the Myrmex agent team in plain language — @-mentions, prompt templates, slash commands, attachments, a message queue, history recall, and downloadable artifacts.

import { CardGroup, Card, Note, Steps, Step, Frame, Tip } from '@mintlify/components';

The chat panel is where you put the AI to work. Type what you want in natural language and **Centurion**, the orchestrator, routes your request to the right specialist. To be precise, use `@`-mentions to call a specific agent or point to a specific resource.

## Scope Requests with `@`-mentions

Type `@` to bring up a picker. You can mention two kinds of things, and each one appears in your message as a colored chip so the agents know exactly what you mean.

<CardGroup cols={2}>
  <Card title="Mention an agent" icon="at">
    Route the request to a specialist — for example `@Perseus` to run a host
    check or `@Scribe` to write a report. Skip it and Centurion picks for you.
  </Card>

  <Card title="Mention a resource" icon="tag">
    Scope the request to a specific **device, integration, report, alert,
    folder, or file** so the agents act on exactly the right asset.
  </Card>
</CardGroup>

<Tip>
  See the full agent roster in the [Multi-Agent System](/documentation/multi-agent-system/overview) and how mentions fit the model in [Core Concepts](/documentation/get-started/concepts#talking-to-the-agents).
</Tip>

## Start from a Prompt Template

If you're not sure how to phrase a request, pick a predefined template to get a well-formed starting prompt you can edit before sending.

<CardGroup cols={2}>
  <Card title="Audit" icon="clipboard-check">
    Kick off an assessment of a device, integration, or your environment.
  </Card>

  <Card title="Configuration" icon="sliders">
    Change or harden the settings of a connected asset.
  </Card>

  <Card title="Investigation" icon="magnifying-glass">
    Look into an alert, indicator, or suspicious behavior.
  </Card>

  <Card title="Report" icon="file-lines">
    Turn an analysis into a shareable document.
  </Card>
</CardGroup>

## Pull in External Documents

Type `/` to run a slash command. Slash commands let you pull a document straight into the conversation from **SharePoint** or **Google Drive** — for example a `/read-document` command backed by your personal connectors — so the agents can work from a spec, policy, or spreadsheet you already have.

## Attach Files and Images

Attach files directly to a message when you want the agents to analyze something you have on hand — documents (PDF, TXT), images (PNG, JPG), and audio. The content becomes part of the request context.

## Work Faster in the Input Box

<CardGroup cols={2}>
  <Card title="Queue follow-ups" icon="layer-group">
    While an agent is working, line up to **three** more messages — they send
    automatically as the current one finishes.
  </Card>

  <Card title="Recall messages" icon="clock-rotate-left">
    Press `Ctrl+↑` and `Ctrl+↓` to cycle through messages you sent earlier and
    reuse them.
  </Card>

  <Card title="Autocomplete words" icon="keyboard">
    Press `Tab` to accept context-aware suggestions for commands and asset names
    as you type.
  </Card>

  <Card title="Track usage" icon="gauge">
    A usage indicator shows how much of the conversation's large context window
    (on the order of one million tokens) you've used.
  </Card>
</CardGroup>

## Act on a Response

Every agent message carries its own actions:

* **Copy** the message text.
* **Rate it** with thumbs up or thumbs down to send feedback that improves the agents.
* **Send to Clipboard** to collect a table, JSON, or code block in the [Clipboard tab](/documentation/ai-workspace/clipboard-and-usage) for editing and reuse.

When an agent produces a document, download the artifact directly from chat as **PDF, DOCX, XLSX, or ZIP**.

## Manage Your Conversations

Start a **New Chat** to pivot to a fresh topic, **Favorite** a conversation to pin it, and open **History** to return to any past thread.

<Note>
  As the agents work, results open as tabs in the Workspace panel on the right,
  so you never lose your place. Watch agents reason and act in [Live reasoning &
  operational panels](/documentation/ai-workspace/live-reasoning).
</Note>