> ## Documentation Index
> Fetch the complete documentation index at: https://docs.typewise.app/llms.txt
> Use this file to discover all available pages before exploring further.
# Chat Configuration
> Customize behavior, appearance, and installation of the chat widget
The Chat Configuration section lets you customize how Typewise AI handles live
chat interactions, including behavior, handoff rules, appearance, and
installation. Setting this up properly ensures smooth communication, consistent
branding, and seamless transitions between AI and Human Agents.
## Guidance
Use the [Guidance](https://platform.typewise.app/channels/chat/guidance) section
to define how Typewise AI should behave during chat interactions and under which
conditions a human should take over.
### Channel Guidelines
Describe what Typewise AI should always do when managing chat tickets. For
example:
* Greet the end customer politely and introduce itself.
* Keep replies short and conversational.
* Use emojis sparingly, only when appropriate.
* Ask clarifying questions if the end customer's request isn't clear.
* Always summarize the resolution before closing the chat.
### Handoff Conditions
Define when the chat should be handed over to a Human Agent. Common examples
include:
* The end customer expresses frustration or asks for a supervisor.
* Sensitive topics like billing, refunds, or legal concerns arise.
### Automation
Fine-tune how your chat behaves automatically with timing and review rules.
* **Automatically follow up after:**
Send a follow-up message after a specified number of minutes/hours/days
without receiving a reply from the end customer. The timer starts after the
last customer message.
When triggered, Typewise AI generates a contextual follow-up message based on
the conversation history, following your channel guidelines and company voice.
For example, the AI might send something like: "Hi there! We just wanted to
check in and see how things are going."
If auto-close is also enabled, the follow-up message may mention when the chat
will close due to inactivity. If "Require human review" is enabled, the
follow-up message waits for approval before being sent.
* **Automatically close tickets after:**
Automatically close chat tickets after a specified number of
minutes/hours/days without a reply from the end customer. The timer starts
from the last customer message.
When the timer expires, the ticket is automatically closed and the end
customer sees that the conversation was closed due to inactivity. If follow-up
is enabled with a shorter timer, the follow-up message will be sent before the
ticket closes.
* **Automatically switch to email channel after:**
Switch communication from chat to email after a specified number of
minutes/hours/days without a Human Agent reply. This setting only triggers
when a handoff to a human agent is pending.
When the timer expires, the AI explains that all team members are currently
busy and offers to continue the conversation via email. It asks for the end
customer's email address (if not already known), and once confirmed, creates a
support ticket with the full conversation history so a Human Agent can follow
up via email when available.
This ensures end customers aren't left waiting indefinitely when chat agents
are unavailable, and that every inquiry is captured and handled.
For flows where the AI should move to email as soon as details are collected
(instead of waiting for this timeout), add the **Switch to email & handoff**
system action in Specialist instructions, global guidelines, or channel
guidelines. See
[System actions](/documentation/actions/system-actions#switch-to-email-%26-handoff).
* **Require human review before sending a message:**
Enable this option if you want every AI-generated message to be reviewed by a
human before being sent to the end customer.
When enabled, all AI-generated messages (including follow-ups) go to a pending
state. Human Agents see the drafted message in their inbox and can approve,
edit, or reject it before sending. This provides oversight for sensitive or
important conversations where you want tighter control over what the AI sends.
***
## Appearance
In [Appearance](https://platform.typewise.app/channels/chat/appearance), you can
configure how your chat widget looks and feels to end customers.
Make sure the appearance matches your company’s branding and creates a welcoming
experience.
### What you can configure:
* **Color & Branding:** Set your brand colors, chat logo, and chat title.
* **Widget Details:** Adjust widget position, icon type, and shape.
* **Greetings:** Customize the welcome message to match your brand voice, as
well as a list of suggested conversation starters the end customer can choose.
***
## Installation
The [Installation](https://platform.typewise.app/channels/chat/installation)
section provides everything you need to integrate the chat widget into your
website or app.
You’ll find:
* **Quick start:** How to embed the chat widget on your site.
* **Domain allowlist:** The domains where your chat widget can load, including
their subdomains. While no domains are configured, the widget loads on any
domain. Add your domains to restrict it. The check relies on the browser's
`Origin` header, so requests that arrive without one, such as server-side
calls, aren't blocked.
* **Identity verification:** Sign the variables your chat widget sends so
Typewise treats them as verified on actions and lookups. Required only if
a chat action or lookup uses the **Trusted only** check on any input. See
[Chat identity verification](/guides/channels/chat-identity-verification)
and [Context Variables](/documentation/settings/context-variables).
To get your chat widget running, complete the **Quick start** and configure
your **Domain allowlist**. Identity verification is optional unless any chat
action or lookup uses the **Trusted only** check. Configure it before
go-live in that case.
Always add the chat widget script directly to your website's HTML source code.
Do not use tag managers (e.g. Google Tag Manager, Adobe Launch, Tealium, Segment) to inject the script. Ad blockers and privacy tools commonly block tag manager requests, which silently prevents the widget from loading for affected visitors.
If your website enforces a Content Security Policy, you also need to allow Typewise domains in your CSP headers. See [Configure CSP for the chat widget](/guides/channels/content-security-policy).
Once installed, your Typewise chat will be fully operational and ready to engage
visitors, answer questions, and escalate seamlessly when needed.
## See also
* [Email Configuration](/documentation/behavior/email): Configure email channel
behavior
* [Specialists](/documentation/behavior/specialists): Create topic-specific AI
behavior
* [Company configuration](/documentation/behavior/company): Set global
guidelines
* [Configure CSP for the chat widget](/guides/channels/content-security-policy):
Required CSP directives for sites with security headers