Welcome to InfraVoice
InfraVoice turns any language and speech model into a production voice agent. Build agents that talk on the phone, pull answers from your knowledge, call your tools, and hand off to humans — all from one dashboard.
Bring your own LLM, ASR, and TTS providers, connect a phone number, and go live. There are no models to host and no real-time infrastructure to run: configure an agent and start taking calls.
The products
Voice Agents
Create AI agents with their own prompt, knowledge, and tools — mix any LLM, ASR, and TTS provider.
Telephony
Take and place real phone calls over Twilio — inbound and outbound, low-latency.
Human Agent Panel
Hand a live call from the AI to a human agent and back — with full context and per-segment transcripts.
Knowledge Base
Ground answers in your own content — workspace memory, runtime KB, and per-session context.
Custom Tools
Let the agent call your APIs mid-conversation to look things up and take real actions.
Speech-to-Text API
Transcribe audio with an sk- API key — batch or real-time streaming, billed per audio-second.
Analytics & QA
Every call transcribed, summarized, and scored — post-call analysis and QA rubrics out of the box.
How it works
- 1
Create an agent
Give it a system prompt, pick your LLM / ASR / TTS providers, and attach a knowledge base or tools.
- 2
Connect a channel
Test it in the browser Voice Test, or bind a phone number for inbound and outbound calls.
- 3
Go live & measure
Route callers to human agents when needed, and review transcripts, summaries, and QA scores for every call.
Start here
Quick Start
Make your first voice agent answer a call in minutes.
API Reference
Grab your organization key and drive the whole platform over HTTP.
User Guide
The full walkthrough — build an agent, connect Twilio, go live.
Voice Test
Talk to any agent live in your browser — no phone number needed.