“Uh, which cable?”
Vapi for
FaceTime
AI video agents that answer and place FaceTime calls, in a few lines of code.

Voice agents are blind.
Chert isn’t.
When the problem is visual, another question rarely helps. A video agent can use the caller’s camera context to understand what they are showing.
“The blue cable, second from the left.”
Three steps to a live FaceTime agent.
You are a calm remote support agent. Ask what the caller sees, then guide one step at a time.
Configure the prompt
Define the job, instructions, model, voice, and behavior before a call is ever placed.
Choose the persona
Select the avatar and framing that fit the experience you want to deliver.
Deploy with a line
Publish the assistant and assign it to a provisioned FaceTime line.
Built for work
that needs eyes.
Use video when a caller needs to show the problem, the environment, or the next step, not describe it from memory.
ChertWhat to know
before you call.
Is the FaceTime API generally available?
Not yet. Chert is onboarding the FaceTime product through a controlled private preview while production media and line readiness are validated.
Can I test an agent without placing a FaceTime call?
Yes. The browser preview is designed for checking the prompt, voice, microphone, interruption behavior, avatar, and cleanup before publishing.
Does it support inbound and outbound calls?
The control plane supports bounded inbound and outbound test workflows. Live execution is provisioned and explicitly authorized; accepting mode is not enabled by default.
What can I configure?
You can configure the assistant instructions, realtime model, voice, avatar, framing, publishing state, and managed line assignment.
Build an agent people
can actually show things to.
Try the controlled preview or talk with us about your first video-agent workflow.