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

# Agent Hub

> The Agent Hub is where the agent asks you for approval, asks questions, and asks for business context during a pentest.

Sometimes the agent needs a person during a pentest. It may want approval before a risky action, or it may need an answer that it cannot find in the application. The Agent Hub collects these requests in one place. Each request is called an interaction.

Open **Agent Hub** in the sidebar. The number next to it shows how many interactions are waiting for you.

## Types of interactions

**1. Approval required**

The agent wants to do something sensitive and waits for your decision. See [Human approval](https://docs.noscope.com/agent-hub/human-approval.md).

**2. Agent question**

The agent has a question about the application. For example, which role a test account has, or how a workflow is supposed to work.

**3. Context needed**

The agent needs business context to judge a result. For example, if a user is allowed to see another team's data.

Each interaction shows a title, the type of attack where relevant, a screenshot of the page, and the agent's reasoning. Approval requests also show a risk level, such as High or Critical.

## Respond to an interaction

* **Approval required**: click **Approve** to let the agent go ahead. You can add a note first; the button then changes to **Approve with feedback**. Click **Reject** to stop the action. When you reject, you can pick a quick reason (for example "Out of scope" or "Too aggressive: lower risk tier") and tell the agent what to do instead.
* **Agent question**: type your answer and click **Submit**.
* **Context needed**: type the context and click **Submit**.
* Click **Ignore** if you do not want to answer. The agent continues without your input.

Only the sub-agent that asked waits for you. The rest of the pentest keeps running.

## Active and Closed

* **Active** lists interactions that are waiting for you.
* **Closed** lists interactions that are done: Approved, Rejected, Answered, Context provided, Ignored, or Expired. Click **View agent reasoning** to see the request and your response.

If nobody responds in time, the interaction expires and the agent treats it as rejected. When a scan ends, its open interactions close automatically and show "Auto-closed when scan ended".

## Find an interaction

* Use the application menu at the top to show **All applications** or one application.
* Use **Search interactions** to search by text.
* Use the **Type** and **Status** filters to narrow the list.

## Get notified

NoScope sends an email when the agent needs your input. You can also get these alerts in Slack. See [Slack notifications](https://docs.noscope.com/integrations/slack-notifications.md).

> [!NOTE]
>
> The agent only sends interactions when [human approval](https://docs.noscope.com/agent-hub/human-approval.md) is turned on for the scan. Human approval is not available for every organization yet. If the switch shows a Locked badge, contact us.


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