Bring your own models.Change them any time.
A voice agent is three skills: hearing, thinking and speaking. Choose a provider for each, paste your key, and switch whenever something better arrives.
What you get
AI models
Leading providers
OpenAI, Google, Microsoft Azure, Amazon Bedrock, Groq, Deepgram, ElevenLabs, Cartesia, Fish Audio and many more.
Guided setup
Every provider has a short plain-language guide with a link straight to where its key is created.
Always-current model lists
Press one button to pull the models a provider offers today, so new releases show up without waiting for an update.
Speech-to-speech
Use realtime models that listen and speak directly when the lowest delay matters most.
Backup voice
Add a second voice provider so a call keeps talking if the first one has trouble.
Your keys, your bill
You pay the providers directly, and call cost is shown per call in your own currency.
How it works
4 steps
Open Models
Choose a provider for each skill
Paste the key from the provider
Pick a model and save
Questions
Which providers are supported?
Speech, language and voice providers including OpenAI, Google, Azure, Amazon Bedrock, Groq, Deepgram, ElevenLabs, Cartesia, Fish Audio, Sarvam and more. The list grows regularly.
Are my keys safe?
Keys are stored for your organization only and are shown masked after saving.
Can different agents use different models?
Yes. The organization default applies everywhere, and any agent can override it.
Keep exploring
Hear it on your own line.
Create a workspace, describe the calls you want handled in plain words, and talk to the agent in your browser first.