Overview#
CallFluence is the voice AI platform for a direct line to your customers: inbound calls are handled by an AI assistant, and website visitors become leads through the callback widget, meeting requests, or chat — fully multi-tenant, all through one API.
This documentation describes what is actually implemented. Endpoints that are only planned are marked as "planned" — nothing is documented here that does not run.
The platform at a glance#
| Building block | Purpose | Service |
|---|---|---|
| Console API | Control plane: assistants, phone numbers, calls, leads, WebWoW_Audio | console-api (port 5810) |
| Voice Runtime | Data plane: receives telephony webhooks and drives the conversation | voice-runtime (port 5811) |
| WebWoW_Audio | Audio touchpoints along the customer journey: scripts, TTS, voice-to-text | part of the Console API |
| Widget | Callback, meeting, and chat leads from any website | callfluence-widget.js |
Core principles#
- One wire, two ports. Telephony carriers (TelephonyPort) and voice AI (VoiceAgentPort) are separate, swappable contracts. No vendor lock-in.
- Tenant isolation is enforced, not a convention. Every data access goes through a tenant scope that forces the tenant ID into every query.
- EU AI Act Art. 50 built in. Every AI-led call deterministically starts with the disclosure that an AI is speaking — it cannot be switched off.
- AI first. Speech models (Cartesia Sonic, Ink, Grok, Whisper) are first-class building blocks, not add-ons.
Start here#
- Quickstart: your first API call in two minutes.
- WebWoW_Audio: audio touchpoints for phone, web, and chat.
- API Reference: every endpoint in detail.