Voice agents that
show their work.
Most voice agents are a black box: you hear an answer and take it on faith. With Remy, you build on a best-in-class voice pipeline without wiring it together yourself.
Every part best-in-class. All of it wired for you.
Build with one realtime model end to end, or piece together a best-in-class model at every stage. Swap any part, any time.
Plus live captions, results that render on screen, and swapping the model or voice mid-call.
157 voices · 20 speech models · metered per minute or per million characters
Talk to Cove.
Cove is a demo, not a product: a made-up smart-home audio brand we built with Remy to show what the platform can do. Everything it does is real, though. Talk to it like a customer, and once you verify your number by text, it pulls up your account and takes real action, like rescheduling a technician visit or placing a callback.
Hi, I'm Cove's assistant.
Ask about an order, how something works, or your account.
Or call it for real.
The same agent you just talked to, now live on a real phone call.
US number · standard call rates apply.
Priced by the unit, metered to the second.
Every voice model on the platform, at its real per-unit rate. Compose any pipeline and pay for exactly what runs, billed per second of audio and per token or character of text. No bundles, no seat minimums.
One model listens, thinks, and speaks. Billed per minute of conversation.
Transcription for composed pipelines. Billed per minute of audio.
Voices for composed pipelines. Billed per character or per token.
Transport, carriage, and numbers. The managed layer both architectures run on.
List prices in USD. Realtime rates are approximate: the per-token rate under each model is what actually meters usage. Your bill is the sum of exactly what each call uses. Volume and committed-use pricing available.