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

# AI Assistant

> Ask questions about a workspace's monitoring data in a conversation, start from an Agent Skill, and check the data behind each answer.

The AI Assistant answers questions about the current workspace in a conversation. It reads the workspace's monitoring data, such as visibility, competitors, sources, sentiment, content opportunities, and connected Measure integrations, and answers with text, tables, and charts.

<Note>
  The AI Assistant is available on every plan and does not consume credits.
</Note>

## Who can use it

Every team member with access to a workspace can use the Assistant in that workspace, including Viewers in the workspaces assigned to them. Conversations are private: each member sees only their own conversations.

## Ask a question

1. Open the workspace and select **Assistant** at the top of the sidebar. Select **Dashboard** to return to the dashboard pages.
2. Type a question, or select an Agent Skill to fill in a ready-made request that you can edit before sending.
3. While the Assistant works, the answer shows the data checks it runs. Expand the details to see them.

The Assistant answers in the language of your message. It may say that some data checks are unavailable: the answer can then be incomplete, so send the request again if you need the missing data.

## Analysis scope

The Assistant does not use the dashboard filter bar. It uses the scope you state in the conversation: period, providers, countries, topics, tags, funnel stages, or personas. Without one, it analyzes all monitored data over the last 30 days.

State the scope in your question when you compare periods, providers, or markets, and check that the answer restates it.

## Agent Skills

When a conversation is empty, **Agent Skills** lists ready-made analyses grouped by category: **Popular**, **Visibility**, **Sources**, **Content**, **Site & Impact**, **Reputation**, and **Help**. Some skills also link to the dashboard pages they rely on, so you can check the underlying data.

## Manage conversations

Select **New conversation** to start a new topic. From the conversation menu, you can **Rename**, **Pin**, **Archive**, or **Restore** a conversation. Pinned conversations stay at the top of the list; archived conversations appear under **Archived**.

| Limit | Value |
| - | - |
| Active conversations per member and workspace | 50. Archive a conversation to start another one. |
| Messages per conversation | 100 |
| Characters per message | 2,000 |
| Messages taken into account | The last 10 messages of the conversation |

When an answer reaches its length limit, select **Continue response**.

## Rate an answer

Select **Helpful** or **Not helpful** under an answer. A negative rating lets you say what should be improved: incorrect data or numbers, filters or scope ignored, an unhelpful recommendation, missing context, an off-topic response, or another reason.

## Verify important answers

The Assistant answers from your monitored data and can make mistakes. Before acting on a figure, open the dashboard page that holds it, such as [Overview](/documentation/cockpit/performance-dashboard) or [Response Analysis](/documentation/monitor/analyzing-answers), and compare the same scope. An answer describes observed data; it does not prove that an action caused a change.

## Related pages

* [Core concepts](/documentation/get-started/core-concepts)
* [Viewers](/documentation/workspace/viewers)
* [MCP server](/mcp/introduction), to query the same data from an external AI assistant


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