All projects
Voice AIIn productionSince 2025

Echo

Real-time voice agents for natural, reliable phone conversations.A low-latency voice stack for phone and in-app voice agents. It streams speech in and out, lets callers interrupt naturally and connects to business systems during the call.

echo · call c_20931 · Clinic 12Example trace
  1. call connected+81 90-••••-•218 · ja-JP
  2. caller: “I’d like to move my cleaning to next week”
  3. ✓ reservations.findThu 10:30 · cleaning
  4. ✓ reservations.availability3 slots offered
  5. ✓ reservations.rescheduleTue 14:15 confirmed
  6. latency p50612 ms
  7. call ended · resolved1m 48s · LINE confirmation sent
median voice-to-voice latency
~650 ms
median voice-to-voice latency
calls fully automated (dental group)
52%
calls fully automated (dental group)
missed-call rate (dental group)
20% → 3%
missed-call rate (dental group)
languages, Japanese first
6
languages, Japanese first

Overview

Why we built Echo.

On a phone call, every hundred milliseconds counts. Echo streams recognition, reasoning and speech synthesis in parallel, predicts the end of a caller’s turn, and starts speaking as soon as it has something useful to say. Callers can interrupt at any time, as they would with a person.

Echo uses Relay for actions, so a voice agent can check availability, book appointments or look up orders mid-conversation, with the same guardrails and traces as any other agent.

Capabilities

Core capabilities.

01

Streaming pipeline

Speech recognition, reasoning and synthesis streamed end-to-end for fast responses.

02

Natural turn-taking

End-of-turn prediction, barge-in and backchannel handling.

03

Telephony built in

SIP trunking, warm transfer to staff, call recording and DTMF fallback.

04

Safety boundaries

Configurable escalation rules, with urgent topics detected and routed immediately.

Architecture

System architecture.

A simplified view of the principal layers and components.

Built with

PythonTypeScriptWebSocketsWebRTCSIPRedisAWS
  1. Telephony

    SIP trunkWebRTCCall routingRecording
  2. Speech

    Streaming STTTurn detectionStreaming TTSBarge-in
  3. Agent

    Dialogue stateRelay toolsEscalation rules
  4. Insights

    TranscriptsOutcomesQA samplingDashboards

Contact

Have an AI product in mind?

Share what you would like to build, and we will help turn the idea into a production-ready product.

No technical specification required.

What happens next

  1. 1

    Share your idea

    Describe the problem, the users and the outcome you need.

  2. 2

    Discovery call

    We review your requirements together and clarify scope and constraints.

  3. 3

    Proposal

    You receive a clear scope, technical approach and delivery plan.