Skip to main content

Pricing & billing

Voice Agents are billed per second of call duration, plus tool and Knowledge Base call charges, plus call connection charges from your telephony provider.

What the per-second rate covers​

The per-second voice rate covers everything inside the voice runtime:

  • LLM usage for the reasoning step
  • STT transcription
  • TTS synthesis
  • Infrastructure base cost — the dedicated voice path
  • Recording charges — separate per-second rates based on your retention requirements

It does not cover:

  • Tool calls invoked during the call — billed under the normal Tool pricing.
  • Knowledge Base calls — billed under the normal KB pricing.
  • Connection charges from your telephony provider (Twilio, WhatsApp, etc.) — these are billed by the provider directly.

How the meter works​

  • Duration is measured from call start to call end.
  • Partial seconds are rounded up.
  • Credit deduction happens at call completion, not in real time.

The per-second credit rate is configured at the platform level and is consistent across voice agents in the workspace.

BYOK vs. platform-managed​

You can bring your own key (BYOK) for the LLM, STT, and TTS providers. The cost effect:

  • When you select the SimplAI API key for a provider, that provider's usage is included in the per-second voice rate. You'll see a small + $0.00X / min cost label next to the SimplAI key option in the key selector.
  • When you select a user-owned key, that provider's usage is not included in the per-second voice rate — it bills directly against your account with the provider. The selector shows no platform cost label for user-owned keys.

See BYOK for how to manage keys.

Usage dashboards​

The workspace usage view breaks voice usage into:

  • Total voice seconds used — over the selected period.
  • Credits consumed — for the per-second voice metering, plus tool/KB charges attributable to voice runs.
  • Platform-managed vs. BYOK — a split so you can see how much usage went through SimplAI-managed keys vs. your own.

Realtime API mode​

The OpenAI Realtime API is billed differently than standard API calls (token + audio second pricing on the OpenAI side). On the SimplAI side, the per-second voice metering still applies for the platform infrastructure layer. If you are using BYOK for the Realtime API LLM, the OpenAI portion bills against your OpenAI account directly. Check with your GTM contact if you need a custom credit rate for Realtime API sessions.

Failed and partial calls​

  • A call that fails before the runtime starts (e.g., a misrouted phone number, missing publish) is not metered.
  • A call that starts and then fails mid-conversation is metered for the time the runtime was active. The status in Run history is marked partial or failed with the duration that was consumed.