Vapi → toolkit-voice

Vapi splits your bill across 5 line-items: platform, LLM, STT, TTS, Twilio. We collapse all five. Same call flow, one invoice.

Your bill, before vs after

Vapi (typical)

Platform fee$0.05/min

LLM (e.g. GPT-4)$0.06/min

Deepgram STT$0.02/min

ElevenLabs TTS$0.06/min

Twilio carrier$0.014/min

$0.21/min

toolkit-voice

Inference$0.04/min

STT + TTS bundled—

Carrier bundled—

Phone number$1/mo flat

Audio retention0 bytes

$0.04/min

Migrate your assistant config

Vapi config maps almost 1:1. The codemod handles 90% of typical configs.

bash

npx @toolkit-llm/migrate-vapi ./vapi-config.json

# Reads your Vapi assistant export, generates:
# ✓ toolkit-config.json
# ✓ migration-notes.md (lists anything that needs manual review)
# ✓ cost-comparison.md (honest before/after at your historical volume)

Vapi config (before):

JSON

{
  "model": { "provider": "openai", "model": "gpt-4-turbo" },
  "voice": { "provider": "11labs", "voiceId": "rachel" },
  "transcriber": { "provider": "deepgram", "model": "nova-2" },
  "firstMessage": "Hello, this is Tony at DC Motorcars.",
  "phoneNumber": { "twilioPhoneNumber": "+14155551234" }
}

toolkit-voice config (after):

JSON

{
  "model": "toolkit-voice",
  "persona": "tony",
  "voice_id": "VARM1",
  "first_message": "Hello, this is Tony at DC Motorcars.",
  "phone_number": "+14155551234"
}

What's the same

Webhook URL pattern (call.started, call.ended, transcript.partial)
Function calling / tool calls — same JSON-schema spec
Phone number management (port your Twilio numbers, no carrier change)
First-message + system prompt + voice configuration
Recording / transcript export (we add: 0-byte retention by default, opt-in storage if you need it)

What's different

No multi-vendor split — one provider, one invoice, one SLA
Per-customer LoRA fine-tunes available Day 270 (Vapi has none)
Briefing context: 4K tokens of fresh world knowledge prefix-cached, free
Pathway YAML — declarative call flows in version control, code-reviewable
ToS anti-clone clause: we never train on your audio, never sell to your competitors

We'll do it for you.

For founding-beta customers, we run the codemod against your Vapi config, send you the diff, and walk you through it on Slack. Most migrations finish in under 30 minutes.

Apply for founding beta →