Build a production voice agent in minutes, not months.

Compose your own STT, LLM, and TTS providers, or use our recommended defaults. Write a system prompt, and Wakili exposes a low-latency voice agent via a single webhook or API endpoint. No infra to manage.

Live call · agent_9f3d2streaming
STT: DeepgramLLM: ClaudeTTS: ElevenLabs

How it works

Four steps between an idea and a live voice agent.

  1. 01

    Pick your agent's brain

    Choose an LLM from our catalog and write the system prompt that defines it.

  2. 02

    Pick its ears and voice

    Choose a speech-to-text and text-to-speech provider — no API keys to manage.

  3. 03

    Test it live

    Talk to your agent right in the dashboard before it ever touches production.

  4. 04

    Connect your app

    Generate an API key, start a call, and stream audio over a WebSocket.

Your STT

Deepgram, AssemblyAI…

Your LLM

GPT, Claude, Gemini…

Your TTS

ElevenLabs, Cartesia…

Wakili

orchestration + low latency

Webhook / API

drop into any app

Your App

Build the exact agent you’d build in the dashboard

Swap the brain, ears, and voice below — the same picker you get in the app — and watch the per-minute rate update live, side by side with the same stack on Vapi.

Everything you need to ship a voice agent

Wakili handles orchestration, latency, and billing so you can focus on the agent itself.

Built-in provider catalog

Pick from our supported LLM, speech-to-text, and text-to-speech providers per agent — no accounts or API keys of your own to manage.

Live cost estimate

See the projected $/minute for a provider combination before you save the agent, right in the settings form.

One dashboard, every setting

Configure the brain, ears, and voice, test calls, and review usage in a single place.

Test calls in the browser

Talk to your agent from the dashboard's Call tab before wiring it into your product.

WebSocket voice protocol

Start a call over REST, then stream mic audio and receive transcripts, replies, and reply audio over a single WebSocket.

Non-realtime endpoints too

One-shot LLM completions (with schema-validated JSON), file transcription with diarization, and standalone speech synthesis — same key, same balance, no second provider.

Signed call.ended webhooks

Register an endpoint and get an HMAC-signed POST when a call ends, with a link to the transcript — no polling.

No subscription

Pay-as-you-go balance

Top up your account once, then every call debits a flat blended rate — the sum of your agent's LLM, STT, and TTS component costs — from your balance.

Per-call cost & metrics

Every call exposes an exact cost breakdown (STT seconds, LLM tokens, TTS characters) plus talk-ratio, word count, and turn count.

Verbatim call history

Every call is recorded with a frozen, verbatim transcript, audio playback, and its exact cost, searchable from the Calls tab.

A real API, not a black box

Generate an API key from an agent’s Connect tab, start a call over REST, then open the returned single-use WebSocket to stream mic audio in and hear your agent talk back. Subscribe to a signed call.ended webhook so you never have to poll for the transcript.

start-call.ts
// 1. Start a call — returns a single-use ws_url
const call = await fetch(`${API_URL}/agents/${agentId}/calls`, {
  method: "POST",
  headers: { "Authorization": `Bearer ${apiKey}` },
}).then(r => r.json());

// 2. Open the WebSocket and start talking
const ws = new WebSocket(call.ws_url);
ws.send(micAudioChunk);
ws.send(JSON.stringify({ type: "end_turn" }));

// 3. Receive JSON events + raw reply-audio frames
{
  "type": "transcript",
  "text": "Can I move my appointment to Friday?"
}

Built for real voice products

Customer support agents

Resolve tier-1 tickets by voice, escalate to humans with full context when needed.

Sales & lead qualification

Qualify inbound calls automatically and route hot leads straight to your reps.

Appointment booking

Let callers schedule, reschedule, and confirm appointments without a human in the loop.

Voice-enabled products

Add a voice interface to your own app without building an STT/LLM/TTS pipeline from scratch.

No plans, no tiers — just a balance

Create an account free. Top up whenever you like, and every call debits one blended per-minute rate — the sum of your agent’s LLM, STT, and TTS component costs.

Top up your balance

Buy a top-up pack from the Billing tab — some packs include a bonus on top of the amount you pay.

One rate per call

No separate STT, LLM, and TTS line items to reconcile — one $/min figure, shown before you save an agent.

Watch it in real time

Balance and transaction history are live in the dashboard — never a surprise invoice.

Get started

Ship your first voice agent today

Free account, no subscription. Top up when you're ready to make your first call.