VOICE API
Bring your agent. We'll put it on a call.
Teler gives your code a phone line - place and receive calls, stream audio both ways while the call is live, and keep your own speech and language models. Numbers in 50+ countries.
How it works
One call, start to finish.
The call connects
Your code dials out, or someone rings your number. Teler asks your flow_url what to do with it.
The audio opens up
The caller’s voice arrives at your server and your agent’s voice goes back, both directions, for as long as the call runs.
Your agent works out what to say
Speech to text, your model, text to speech. This part stays yours.
You keep control until the call ends
Interrupt, transfer to a person, record, hang up. Every event arrives as a signed message you can replay.
Built for AI voice agents
Someone will talk over your agent
Sometimes a person has to take over
You need a record you can trust
Someone will talk over your agent
Send one message and everything your agent had queued stops playing immediately. You decide what counts as an interruption - Teler doesn't guess for you.
Sometimes a person has to take over
Hand the call across to a colleague, cold. You control what the caller hears while it happens, and what your colleague hears when they pick up, so they know the situation before they say hello.
You need a record you can trust
Teler sends a signed message for everything that happens on a call, and you can ask for it again later if your service missed it. Recordings and the full event log come back through the API.
What you can build
AI voice agent that answers and dials
Inbound and outbound, on the same number, with the same code.
IVR where the caller just talks
No menu tree, no keypad instructions.
Set different recording rules for different teams
Your users press call in your app, and the call goes out under your number.
Escalation that doesn't drop the call
Your agent takes the first two minutes and a person takes the rest.
Appointment reminders that can reschedule
Confirm the time, or offer three others, without a human on the line.
Verification calls for one-time codes
Read back a code, confirm identity, and hang up.
Works with your favorite tools
Teler connects natively with OpenAI Realtime, Gemini Live, ElevenLabs, VAPI, Deepgram, Ultravox, or any custom model - no lock-in, and you pay your provider directly.
Teams already running on this
What is a voice API?
A voice API lets your software make and receive real phone calls in code, instead of through a phone system. Your application sends a request, the call goes out over the phone network, and your code decides what happens while the call is live.
How is a voice API different from a SIP trunk?
A SIP trunk carries the call. A voice API carries the call and gives your code control over it - what plays, what streams where, when to transfer, when to stop.
Inbound, outbound, or both?
Both, on the same number.
Which models can I use?
Yours. OpenAI, Gemini, ElevenLabs, Deepgram, VAPI, and Ultravox are all in use today, most with a bridge you can clone.
Can I create a SIP trunk in Teler?
Yes, you can create and manage multiple SIP trunks in Teler.
Does it connect to my CRM?
No. Teler has no native CRM or help desk connectors - you get webhooks and an API, and you write the integration. If you want calling that's already inside your CRM, that's FreJun Dialer, not Teler.
How fast is the round trip?
Under 100ms, each way - the time for the caller's audio to reach your server, and for your response to reach the caller back. Both directions are covered by that same figure, not just one leg of it.
Where can I get numbers?
50+ countries - local, toll-free, and mobile, provisioned through the API.
Your first call is a
quickstart away