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

# Finish AI Chatbot setup

> Create a web channel, configure allowed origins and the widget, install the snippet, and test the visitor experience.

Chatbot signup intentionally stops before creating a live website channel. After you verify the account and sign in, the Chatbot workspace gives you a three-step launch checklist.

## The setup checklist

> **1. Connect an AI model**
>
> Confirm that at least one active model is available for customer conversations.

> **2. Create a web channel**
>
> Define where the Chatbot may run and how the channel should behave.

> **3. Install and test**
>
> Copy the generated widget snippet and verify a real visitor conversation.

## 1. Confirm the AI model

Open **AI** in the RAUM dashboard.

If you connected a provider during signup, an active model may already be configured. Otherwise, add the provider credentials and models before creating live Chatbot traffic.

The Chatbot cannot answer customer questions without an active compatible model.

[AI provider setup →](https://docs.raum.am/configuration/ai-provider.md)

## 2. Create a web channel

Open **Channels** and create the website channel.

Configure the **exact origins** where the widget is allowed to operate. For example, production and staging websites should be added deliberately rather than using a broad wildcard simply because wildcards are convenient and security incidents are apparently a hobby.

You can also configure channel-level behavior such as appearance and optional human handoff.

> [!WARNING]
>
> Only add domains you control and intend to use with this widget. Origin checks are part of the public Chatbot security boundary.

## 3. Configure the widget

Open **Widgets** and review:

* **Appearance**
* **Languages**
* **Experience & CSAT**
* **Installation**
* **Preview**

Use Preview to check the visitor experience before embedding the widget broadly.

## 4. Install the widget

Open **Widgets → Installation** and copy the generated installation code.

The generated snippet contains the public widget identifiers and publishable key information required by the loader. Use the current snippet from the dashboard instead of copying an old example from another environment.

Add it to the website origin you configured for the channel.

## 5. Test a conversation

Test the same path a real visitor will use.

Verify that:

* the widget loads only on an allowed origin;
* a new conversation can be created;
* RAUM answers using the expected knowledge;
* the correct language and appearance are used;
* CSAT behavior matches your configuration;
* any public lookup or action follows the configured approval and confirmation rules; and
* human handoff behaves as expected.

## 6. Review operations

After chats begin arriving, open **Conversations** to inspect transcripts and delivery behavior.

Use **Requests** for conversations that require human follow-up and **Topics** to review classification and visitor-data mappings.

## If you purchased both products

Your Assistant and Chatbot can reuse compatible organization-level knowledge and AI configuration, but the public widget has its own security boundary.

Internal Zendesk resources and connections are not automatically exposed to website visitors. Approve Chatbot-facing knowledge and public connection access deliberately.

> **[AI Chatbot overview](https://docs.raum.am/chatbot/overview.md)**
>
> Understand the customer-facing Chatbot product and its boundaries.

> **[Channels and widgets](https://docs.raum.am/chatbot/widgets.md)**
>
> Configure allowed origins, appearance, languages, installation, and preview.


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