Website widget settings
Entry point settings decide where the widget may run and how visitors talk to the agent. They are separate from the widget design, which only controls appearance.
To open them, go to Channels, open the website channel and select Configure & install. Settings are on the General and Conversation tabs. Select Save entry point to apply your changes; Discard drops them. Saving keeps the same widget ID, so your embed code does not change.
You need the Owner, Admin or Builder role to change these settings.
General
Section titled “General”| Setting | What it does |
|---|---|
| Entry point name | Shown in the console only. |
| Allowed website origin | The exact origins that may show the widget, up to 10. Use HTTPS, or HTTP on localhost, 127.0.0.1 or a .localhost host. Select Add origin for more. See Allowed origins. |
| Enable this website entry point | Turn off to stop the widget accepting new conversations without deleting the channel. |
| Remove entry point | Deletes this website channel and its installation. The agent and the widget design are kept. |
The Appearance card on the same tab is where you apply a published design revision. Applying takes effect immediately and is not part of Save entry point.
Text and voice
Section titled “Text and voice”The widget supports text chat and, optionally, voice conversations in the browser.
Voice is off for a new website channel unless you selected Allow voice conversations when creating it. To turn it on later:
-
On the channel page, select Edit channel.
-
Select Allow website voice, then Save changes.
-
In the entry point’s Conversation tab, set Conversation modes to Text + voice and select Save entry point.
| Conversation mode | Visitors can |
|---|---|
| Text only | Chat by text. |
| Text + voice | Chat by text and start a voice conversation from the widget. The browser asks for microphone access the first time. |
Turning Allow website voice off switches every website entry point of the channel back to text only.
Voice engine and turn-taking
Section titled “Voice engine and turn-taking”A website voice conversation uses the voice settings of the agent release the channel serves: voice language, speech recognition model, voice, optional realtime model, and timing such as Start replies early. If the agent has no voice settings, Caplo uses its default English speech recognition and voice, and no setup is needed.
You change these on the agent, under Advanced settings then Voice in its configuration, and they reach the website after you publish a release and update the channel. See Voice and speech.
Conversation actions
Section titled “Conversation actions”Conversation actions lists the actions this entry point allows, such as Transfer to a human. An action is available only when the agent release also allows it. For a transfer, choose the Transfer destination from your handoff destinations. See Handoff and end actions.
Customer verification
Section titled “Customer verification”Customer verification lets visitors prove who they are, so the agent’s tools can act on their own account. You set up the verification methods themselves as connections under Settings then Customer identity: Host sign-in (your website’s backend signs a short assertion for a visitor it already knows), Provider sign-in (your OpenID Connect provider in a popup) or Verification code (a one-time code sent to the visitor’s email or phone on file). On the entry point you choose which of them this website offers.
| Customer verification | Behavior |
|---|---|
| Anonymous visitors only | No verification is offered. |
| Visitors may verify who they are | Visitors can chat anonymously and verify when they or a tool need it. |
| Visitors must verify before chatting | Visitors verify before the agent answers. |
When verification is allowed, select Choose sign-in methods and consent for this channel to set:
- Sign-in methods: up to 4 connections, in the order the widget offers them, each with an optional button label. Host sign-in (your own page passing a signed token) runs on its own when your page registers a token provider with the JavaScript API.
- Consent mode: Ask for agreement before the first verification, or Show a notice after verification. Host sign-in only runs automatically with the notice mode.
- Consent version and Consent text: the text visitors agree to, in English and any translations you add. Change the version whenever the text changes; visitors agree to a specific version.
- Previous conversation: Never show earlier conversations, or Show the most recent one after verification.
If the agent’s tools declare verified operations, Verified tool operations lets you raise the verification level an operation needs, require recent authentication, and send your API a customer context token. See Customer-aware tools.
Visitor files during handoff
Section titled “Visitor files during handoff”While a person from your team handles the conversation, visitors can attach files for them. The agent never sees these files, and each file is checked for type and malware.
Select Let visitors send files during a handoff, then choose:
- Accepted files: Images (JPEG, PNG, GIF, WebP) and PDF documents. Both are selected by default.
- Largest file: 1 MB, 5 MB (the default) or 10 MB.