Speak without sending each recording
The continuous voice mode listens and responds within a live browser session. Reviewed voice notes remain available as a separate option.
Features / Real-time voice
Real-time voice
Let your customers say it out loud.
Offer continuous website voice conversations alongside text chat. Visitors can speak, hear the agent and keep the conversation in the same experience.
Get startedThe visitor starts website voice and speaks naturally. The agent uses the same approved knowledge and tools.
Explore this featureThe everyday difference
Make the next step easier.
Choose the speech setup
Use your configured local speech service or opt into OpenAI Realtime where your workspace permits it.
The same business tools
Voice uses the published agent graph and authorized knowledge and business tools. Switching to voice does not grant extra access.
From setup to first conversation
Here’s how to start.
Configure it in your workspace, then test the experience before you publish.
- 1
Configure speech
Set up the voice services and select the allowed voice provider in your workspace.
- 2
Review and publish
Check the agent graph and publish the voice configuration and tool permissions.
- 3
Let visitors choose
Visitors start the voice session and allow their microphone. The audio destination is shown before capture.
A few useful details
Before you get started.
Can a visitor still type?+
Yes. Text chat and voice notes remain available. Visitors choose how they want to communicate.
Does the website make phone calls?+
No. This release supports voice in the browser. Telephone numbers, SIP and outbound calls are not included.
Better together
Explore the rest of your toolkit.
Make room for better conversations
A better conversation starts here.
Create your workspace, connect what you need, and test your agent with the questions your customers actually ask.