Wakili

Migration guide

Moving from Vapi to Wakili

Vapi's advertised rate starts around $0.05/min, but that only covers orchestration. Once you add transcription, an LLM, and a voice engine, most teams land closer to $0.12-$0.26/min across three separate provider bills. Wakili keeps the same idea (bring your own STT, LLM, and TTS) but bills it as one flat platform fee on top of whatever your providers actually charge.

Why teams switch

The sticker price isn't the real price

$0.05/min is the orchestration layer only. Transcription, LLM tokens, and voice generation are billed on top, often across three separate vendor accounts.

Four to six providers to reconcile

Budgeting a call means adding up STT, LLM, TTS, and telephony line items from different dashboards every month.

Still locked into one orchestration bill

You already picked your own providers. Wakili just adds one predictable platform fee instead of a black-box orchestration charge.

Cost, side by side

Figures reflect publicly listed rates and typical stacked costs at the time of writing. Confirm current numbers before you commit.

VapiWakili
Base rate$0.05/min (orchestration only)$0.06/min (all platform fees)
Realistic all-in cost$0.12-$0.26/min$0.11-$0.13/min
Free tierLimited trial credits30 min/mo, free forever
BillingPlatform fee + 3 provider invoicesOne invoice, providers itemized

What maps where

Most Vapi agents map onto Wakili one field at a time. Nothing about your prompt, voice, or model choice has to change.

FieldVapiWakili
Assistant / agentassistant.create()agents.create()
System promptassistant.model.messages[0]systemPrompt
Voiceassistant.voice.voiceIdtts
Transcriberassistant.transcriberstt
Server URLassistant.serverUrlwebhookUrl
create-agent.ts
// Same idea as your Vapi agent config
const agent = await wakili.agents.create({
  stt: "deepgram:nova-3",
  llm: "anthropic:claude-sonnet",
  tts: "elevenlabs:rachel",
  systemPrompt: "Paste your existing prompt here…",
  webhookUrl: "https://api.acme.com/wakili",
});

How the switch works

  1. 0101

    Export your config

    Pull your agent's system prompt, voice ID, LLM model, and webhook URL.

  2. 0202

    Map fields into Wakili

    Same STT, LLM, and TTS choices if you want them. Nothing forces a swap.

  3. 0303

    Run both in parallel

    Point a test number at Wakili and compare transcripts side by side.

  4. 0404

    Cut over your number

    Swap the webhook, retire the old plan. Most teams do this in an afternoon.

Ready when you are

Bring your Vapi agent over

Keep your prompt, your voice, your model. Just change where the bill comes from.

Start BuildingTalk to Us