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

# Assistants

> Configure the conversation brain: instructions, model, knowledge base, and MCP.

<Tip>
  **Assistant** = who the AI **is** and **how it talks**. **Knowledge base** = what it **looks up**. **MCP** = what it **can do** (email, calendar, your API).
</Tip>

## What is it?

The AI profile used in the **AI assistant** graph step (or on-site chat). You set name, instructions, model, and linked resources.

## Configure in the dashboard

**Automations** → **Assistants** tab.

| Field              | Tip                                                                                             |
| ------------------ | ----------------------------------------------------------------------------------------------- |
| **Name**           | Clear for the team (`Store Support`)                                                            |
| **Instructions**   | Short and focused, tone, limits, when to hand off                                               |
| **Model**          | More capable = long context; lighter = fast and cheaper                                         |
| **Knowledge base** | One or more bases, see [Knowledge base](/en/automations-api/como-funciona/base-de-conhecimento) |
| **MCP / HTTP**     | External tools, see [MCP integrations](/en/automations-api/como-funciona/integracoes-mcp)       |

Example instruction:

> You are support for Store ABC. Reply in English, politely and briefly. If you don't know, say you'll get a human. Never invent prices.

## Use in an automation

Typical chatbot flow:

```
Trigger (WhatsApp message) → AI assistant → (optional) Condition → Human or End
```

In the editor, add **AI assistant** step and pick the assistant you created.

## Best practices

| Do                               | Avoid                      |
| -------------------------------- | -------------------------- |
| Short instructions with examples | Huge unfocused prompt      |
| FAQ in small topics              | One unstructured PDF       |
| Path to human                    | Infinite loop with no exit |
| Test real questions              | Ship without testing       |

## v1 API

Assistants are configured **in the dashboard** (**Automations → Assistants**). There is no public `/v1/` endpoint for assistant CRUD in this documentation.

For graphs, events, and runs, use the [v1 API](/en/automations-api/como-funciona/introducao) with an API Key.

## Next steps

* [Knowledge base](/en/automations-api/como-funciona/base-de-conhecimento)
* [MCP integrations](/en/automations-api/como-funciona/integracoes-mcp)
* [Use cases](/en/automations-api/como-funciona/casos-de-uso)