Pay for the management layer. Pick how you handle provider costs.
Most customers use prepaid credits with auto-reload — set a floor, top up automatically when low, one bill, total transparency. Or bring your own provider keys if you prefer direct vendor relationships. Wax charges only for the harness above the model: failover, voice identity across providers, the visual builder, observability, audit-grade governance.
Not ready to commit? Sign in to try a call from your browser.
One price per minute. Fully itemized.
Every conversation minute is broken into its components. LLM, STT, and TTS pass through at cost — 0% markup on provider charges. The flat Wax fee is the only thing Wax marks up.
| LLM | your model, at cost | $X/min |
| STT | your provider, at cost | $Y/min |
| TTS | your provider, at cost | $Z/min |
| Wax | flat fee, every minute | $0.03/min |
| Total | per conversation minute | LLM + STT + TTS + $0.03 |
Every account gets 150 conversation minutes each calendar month — every feature, every provider, no restrictions. After that, prepaid credits.
Prepaid credits
After your free allotment, usage draws from your credit balance. Minimum purchase: $1. Suggested amounts: $25 $100 $500. Opt in to auto-top-up and your balance reloads automatically when it gets low.
Minimum: $1 · No expiry · Opt-in auto-top-up · No subscription
Two ways to handle provider costs.
Choose per component — LLM, STT, and TTS can each be routed differently.
Wax holds the provider account and bills you its cost — no markup, itemized on every minute. One bill, every minute auditable. The same model OpenAI, Anthropic, and Twilio use.
Use your own provider account for any component. That provider bills you directly; Wax charges $0 on that line. Keep your vendor relationship and volume discounts — Wax still handles failover and voice identity.
What the $0.03/min covers
When your primary provider errors or times out, Wax switches that turn to your backup before the agent starts speaking its reply. A hung vendor can add a short wait (about 6 seconds for speech, plus the backup's first audio: up to about 5 seconds on Mistral Voxtral) on that one turn.
Clone your voice on ElevenLabs, Mistral Voxtral, or Cartesia and the backup speaks in the same voice. Deepgram has fixed voices and does not clone, so a switch to it changes the voice.
Build and edit agents in one surface, then watch every turn — latency, cost, and routing decisions itemized per call.
Every minute is attributable to a tenant, agent, and provider. Itemized cost, signed usage events, exportable records.
Cost calculator — example scenarios at 10,000 min/month
Wax flat fee: $0.0300/min on top of provider cost. ElevenLabs list rate: $0.3000/min. Cartesia: $0.0700/min. Deepgram: $0.0600/min. 100% EL passthrough is a cost negative vs EL-direct — Wax adds the flat fee on top.
| Scenario (EL/Cartesia/Deepgram mix) | Minutes/month | Blended $/min (incl. Wax fee) | Total monthly cost | Delta vs 100% EL-direct |
|---|---|---|---|---|
| 100% EL passthrough (BYO-EL) | 10,000 | $0.3300 | $3,300.00 | +$300.00 (more expensive vs EL-direct) |
| 80/10/10 light mix | 10,000 | $0.2830 | $2,830.00 | $-170.00 (cheaper vs EL-direct) |
| 50/30/20 aggressive mix | 10,000 | $0.2130 | $2,130.00 | $-870.00 (cheaper vs EL-direct) |
Interactive calculator available when JavaScript is enabled. Rates as of 2026-05 from public provider pricing pages.