~360ms voice-to-voice · measured in production

Give your product a voiceyour users can just talk to

Build voice agents that answer, call out, and hold a real conversation. One flat monthly price, unlimited minutes — you bring your own model, speech and telephony keys.

Start building free
readyVocra demo · en-US
WebSocket · PCM16 16k
Tap to talk
press space to talk
Tap the orb and just talk

A free demo on our keys. Sign up free to build your own agents with your own providers.

ROUND-TRIP —
Unlimited minutesBring your own keysPhone + web
~360ms
Median voice latency
14
Recognised languages
∞
Minutes, every plan
11
Providers, your choice
Under the hood

One streaming pipeline, your providers

Caller audio
WebRTC · Twilio
Speech to text
Deepgram Nova
Reasoning
Groq · OpenAI · Anthropic · Gemini · DeepSeek · Grok · Kimi · Qwen
Voice out
Deepgram · ElevenLabs

Every stage streams into the next — the agent starts speaking before the model finishes thinking. Interrupt at any moment and the pipeline flushes within one frame.

Bring your own providers & models

GroqOpenAIAnthropicGeminiDeepSeekGrokKimiQwenDeepgramElevenLabsTwilio

Everything to run voice agents in production

Configure the brain and the voice, place or receive real phone calls, and watch every conversation land in one dashboard.

Natural turn-taking

Vocra knows when to listen, when to pause, and when to jump in. Interruptions cut the agent off mid-sentence, the way a real conversation flows.

Sub-second latency

Around 360ms voice-to-voice, measured in production. The reply starts streaming before the model has finished generating the rest of it.

Real phone numbers

Point an agent at a phone number and it dials out, handles the conversation, and hands you a transcript. Connect your own Twilio account.

Bring your own providers

Groq, OpenAI, Anthropic, Gemini, DeepSeek, Grok, Kimi or Qwen power the reasoning; Deepgram or ElevenLabs the voice. Swap providers per agent — your keys, your cost.

Full transcripts

Every call recorded and transcribed with per-turn latency, so you can see exactly where the time went — not just that a call happened.

One flat price

Unlimited minutes on every plan. You pay your model, speech and telephony providers directly at cost — Vocra bills only the platform.

Developer API

Launch a call in one request

A REST endpoint places the call; events and transcript come back over the API once it completes. SDKs are on the roadmap — today it's a plain HTTP call with a bearer key.

Read the docs →
POST /api/calls/outbound
// place an outbound call
await fetch("https://vocra.cloud/api/calls/outbound", {
  method: "POST",
  headers: { Authorization: `Bearer ${VOCRA_KEY}` },
  body: JSON.stringify({
    agentId, to: "+201234567890",
  }),
});

Hear it for yourself

Build your first voice agent free. No credit card — bring your own keys and start talking.

Start building free →
Vocra — Realtime voice for AI agents