LiveSpeech docs.

Build a phone AI agent that answers and places real calls. The full documentation is being written — everything below works today.

Install

# TypeScript
npm install @drawdream/livespeech

# Rust
cargo add livespeech-sdk

Quickstart — three lines

import { LiveSpeechClient } from '@drawdream/livespeech'

const client = new LiveSpeechClient({ apiKey: process.env.LIVESPEECH_KEY })
await client.startSession({ prePrompt: 'You are a scheduling agent.' })

Sessions stream over WebSocket. Interruption control, tool registration, and conversation memory are first-class SDK APIs — the package READMEs cover them until the full guides land here.

In progress

Guides being written: phone number setup, tool calling end to end, interruption handling, multilingual sessions, and the WebSocket protocol reference. If you need one of these now, the console walks you through the working paths.