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

# AI Agent

> The panel beside the dashboard: ask a question about this brand's data in plain words, and open the view that shows the answer.

The AI Agent is a panel beside the dashboard. You type a question about this brand's data, and the answer comes from the answers FullReach AI collected for the open project. Each answer ends with a link to the dashboard view that shows its evidence.

Open it with **AI Agent** in the top bar, or with Ctrl+J (⌘J on a Mac). It stays open while you move between pages, so a link in an answer changes the page beside it and keeps the chat.

The AI Agent answers your questions. It is not the AI bots that read your site, which [Crawler visits](/reference/crawler-visits) counts.

## What it reads

The AI Agent reads the same figures as the dashboard and as the [MCP server](/mcp/introduction). It starts from the page you have open, with its filters. It then reads other views when the question needs them. The line above each answer names the views it read.

The AI Agent reads only the open project. It cannot change the project. When you ask for a change, it says where in the dashboard you make it.

To send a view's data with the question, press **@** and pick the view. To start from a common question, type **/** or pick a quick start in the empty panel.

## Normal and Deep

Each question runs at one of two depths:

| Depth | Use it for |
| - | - |
| Normal | Most questions: a figure, a comparison, a list |
| Deep | A question that asks why something changed, or that joins several views |

A Deep question uses more of the allowance than a Normal one. The Deep choice in the panel shows how many times more. When the deeper model cannot answer, a Normal model answers, and the question uses only a Normal question's share. The answer then shows no **Deep** mark.

## The allowance

Each plan holds an allowance of questions, and each extra project on Advanced adds to it. The usage button in the panel's top bar shows how much of it your organization used in the last 24 hours, the last 7 days and the last 30 days. Each window has its own share of the allowance, so a busy day cannot use up the month. Older questions age out of a window, and room comes back as they do.

A question that no model answered uses none of the allowance. The panel says so and offers **Try again**.

## Chats

Your questions about one project form a chat. **New chat** starts an empty one, and the history button opens your earlier chats about the open project. Only you see your chats.


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