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

# Tools

Tools extend what your agent can do beyond answering questions. Right now there are two: **Human Escalation** and **Lead Capture**. Each is a simple on/off toggle - no complex setup required.

You'll find them under **Tools** in the sidebar.

> **Plan requirement:** Tools are available on the **Builder & Pro** plans. On Free and Starter, the page is visible but the toggles are locked. Head to **Settings → Billing** to upgrade.

> **Access:** You need Editor or Owner access to enable or disable tools. Viewers can't open this page.

***

## Human Escalation

When this tool is enabled, your agent can hand off a conversation to a human team member - either on its own judgment or immediately when a visitor asks to speak to a real person.

### What changes in the chat

The agent gets an explicit escalation rule added to its system prompt. In practice, it will:

* Escalate automatically if it can't find a relevant answer after searching its knowledge base and the topic needs real help
* Escalate immediately if the visitor says something like "I want to talk to a human" or "Can I speak to someone?"
* Send the visitor a short farewell message explaining that a team member will follow up

Once escalated, the AI stops generating responses. The visitor can still type and send messages - they're held for a human to reply to.

> **What it won't do:** The agent won't escalate for greetings, thank-yous, or casual small talk. It also won't escalate when a visitor just wants to leave their contact info for later - that's what **Lead Capture** is for.

### What changes in the inbox

Escalated conversations are flagged with an **"Escalated"** badge in the conversation list and detail header. In an escalated conversation:

* A collapsible **Escalation reason** section shows the brief reason the AI passed off (e.g. "User asked about a refund that requires account access").
* The reply composer switches to **"Reply as human agent"** mode - anything you type is sent as a human message, not an AI response.
* The visitor sees your replies in real time.

You can reply in the inbox to any open escalated conversation. Closing the conversation ends the human takeover session.

### Email notifications

When a conversation is escalated (by the AI), all **workspace owners** receive an email notification. The email includes a direct link to the conversation in the inbox.

Notifications are sent once per escalation event - no repeated emails if more messages come in.

***

## Lead Capture

When this tool is enabled, your agent can show visitors a small contact form - name and email - so they can leave their details for follow-up without needing a live conversation.

### What changes in the chat

The agent gets lead capture rules added to its system prompt. It will offer the form when:

* A visitor asks about pricing, a demo, or anything that warrants a follow-up call
* A visitor wants to be contacted later but isn't asking for a live human chat right now

The contact form appears inline in the conversation - the visitor fills in their name and email and submits. The agent continues the conversation naturally while the form is visible; it won't push the visitor to fill it in again once it's been shown.

### What appears in the inbox

Once a visitor submits their contact details:

* Their **name and email** appear in the conversation header
* A collapsible **Lead capture context** section shows the brief context the AI noted when it triggered the form (e.g. "visitor asked about pricing")
* The lead is stored and deduplicated by email - if the same visitor submits details in multiple conversations, they're linked to the same lead record

### Email notifications

When a visitor submits the lead form, all **workspace owners** receive an email with the visitor's name, email address, and a direct link to the conversation.

***

## How the two tools work together

Human Escalation and Lead Capture serve different intents - and the agent knows the difference:

| Visitor intent                                  | Tool used        |
| ----------------------------------------------- | ---------------- |
| "I need help right now, can I talk to someone?" | Human Escalation |
| "Can someone get back to me about pricing?"     | Lead Capture     |
| "I need a demo scheduled"                       | Lead Capture     |
| "I can't figure this out, I need a human"       | Human Escalation |

You can enable both tools at the same time - the agent will use whichever is appropriate based on what the visitor says.

***

**Next up:** [Operate Daily → Inbox](../en/operate/inbox) - manage conversations, reply as a human, and work with leads.