Answers from your knowledge
The voice agent reads the same uploaded documents, pages, and policies as your chat agent. One set of content keeps both honest.
The loop
What happens in a single spoken turn
A voice conversation is just this loop, repeating, until the visitor has what they came for.
- 01
Visitor speaks
One click, allow the mic, and they start talking. No typing, no phone number, no app to install.
- 02
Agent listens
Speech is understood as it arrives, so the agent is already working on the answer while they finish the sentence.
- 03
Agent acts
Mid-call it can search your knowledge, look something up, save a lead, or open a ticket.
- 04
Agent answers
It speaks the answer back out loud, and the visitor can cut in whenever they want to change direction.
Capabilities
A voice agent that can actually do things
Talking is the easy part. The value is in what the agent can reach while the visitor is still on the call.
Saves what it learns
Leads, tickets, feedback, and booking requests are written into your records mid-conversation, not typed up afterwards.
Runs your connected tools
The same helpers your chat agent uses are available on a call, so it can look up an order or draft a follow-up while the visitor is still speaking.
Writes down every word
Both sides of the call are transcribed as they speak, so the conversation is fully readable the moment it ends.
Handles interruptions
If a visitor talks over the agent, it stops immediately and listens instead of finishing its sentence. That is what makes it feel like a conversation.
Hands over cleanly
When a call needs a person, the agent captures the context and passes it on, so your team is not starting from scratch.
One agent, two ways in
Voice is not a second product to maintain
You do not build a chat agent and then rebuild it for voice. It is the same agent, reached a different way. Fix a wrong answer once and both channels are fixed.
Visitors can switch too. Someone can start by typing and move to speaking when explaining gets tedious.
Shared setup
Chat
Voice
Your instructions and tone
Your knowledge base
Your records and datastores
Your connected tools
Post-conversation review
The embed on your site
When to use it
Where speaking beats typing
Voice is not always the right answer. These are the moments where it clearly is.
Reception and front desk
Visitors arriving on your site with a simple question get an answer spoken back instead of hunting through a help centre.
Hands and eyes are busy
On a shop floor, in a vehicle, or mid-task, speaking is the only practical way to ask a question.
Accessibility
Not everyone can comfortably type a long question. Voice removes that barrier without a separate experience to maintain.
Messy, layered questions
Some problems are far faster to explain out loud than to write down. Voice gets to the real question in one turn.
FAQ
Common questions about voice
Do I need a phone number or a call provider?
No. The voice agent runs inside the browser tab through the same embed as your chat agent. There is no phone line to buy, no dial-in number, and nothing for the visitor to install.
Can a visitor interrupt the agent?
Yes. If the visitor starts speaking while the agent is talking, the agent stops and listens. It keeps what it had already said in the transcript so the record stays accurate.
Do I have to build a second agent for voice?
No. Voice and chat are two ways of reaching the same agent. It uses the same instructions, the same knowledge, and the same actions, so anything you improve applies to both.
What happens to the conversation afterwards?
Everything spoken is transcribed while the call is happening. When the call ends, the transcript is reviewed and turned into a structured record with a summary, an outcome, sentiment, and follow-up actions.
Does the visitor need to give microphone access?
Yes, the browser asks for microphone permission before a voice session starts, and the visitor has to allow it. If they would rather not, they can keep using text chat in the same widget.