voice, answered
Everything you need to deploy, tune and measure your voice agent.
ECHO is a voice-agent platform: deploy a human-sounding agent that answers every call, resolves most on the spot, and hands off the rest to a human with full context. No prompt engineering, no training data.
Everything runs over https://api.echo.ai. The CLI is the single source of truth: echo init, echo connect, echo deploy.
$ npm i -g echo $ echo init $ echo connect zendesk --kb support-docs $ echo deploy ✓ agent live · 1.2s answer time · 92% resolved
The core commands, in the order you'll use them.
| Command | What it does |
|---|---|
echo init | Create an agent and its workspace |
echo connect <source> | Link a knowledge base, docs or CRM |
echo deploy | Go live on a number |
echo transcript <id> | Pull a call transcript with intent & sentiment |
echo handoff <id> | Escalate a live call to a human |
An agent is a set of instructions plus a knowledge base: how it sounds, what it knows, and when it escalates. ECHO drafts all three from a sentence, then you review a few sample calls and ship.
Guardrails define when a call leaves the agent — tone, sensitivity, complexity, or a specific phrase. When a call crosses one, ECHO routes it to a human with the full transcript, intent and sentiment attached, so no one ever starts from zero.
Every call is transcribed in real time with intent, sentiment and resolution status. Search any conversation, replay the audio, and export the data to your warehouse.
Call data is encrypted in transit and at rest, and access is scoped with SSO. We are SOC 2 Type II audited and never train on your data.