Guides for building and operating real-time voice and multiplayer apps on VoiceThere. More sections are rolling out ahead of public launch on voicethere.cloud.

Use the sidebar search to jump to a guide by title, keyword, or topic.

Coding agents can fetch /llms.txt for a machine-readable index of docs, APIs, and npm packages — see LLM agent discovery.

Product announcements and release notes live on the blog.

Get started

LLM agent discovery — How LLM agents discover VoiceThere — /llms.txt, /llms-full.txt, deployed features, on-platform STT and TTS, pricing, OpenAPI, and npm packages.
Quickstart — End-to-end path — login, create a project, build and upload an agent, deploy to cloud, and connect a browser client.
Live showcase demos — Try voice and multiplayer demos on the marketing site — landing chat, showcase catalog, shared world, and how operators configure demo projects with client keys.
Spatial audio showcases — Orbiting sine tone and TTS, spatial soundboard, and proximity room — live demos, spatial-showcase template, DataChannel join protocol, and browser master volume.
Account signup — Create a VoiceThere dashboard account — public signup, operator kill switch, and optional email allowlists.
CLI login — Sign in to VoiceThere from the CLI — browser device approval, personal API keys, and CI-friendly manual login.
CLI command reference — Every voicethere command by subject, with usage, options, an example, and example output.

npm packages

npm packages — Public @voicethere packages for agents, CLI, and clients — plus the open WebRTC SDK they sit on.
@voicethere/agent — Sandboxed agent SDK — defineAgent, speak, data channels, clip playback, mix poses, and bundle build/verify.
@voicethere/cli — voicethere login, init, source sync, build upload/promote/download, deploy --wait, and project settings.
@voicethere/client — Browser, embed, and Node clients — connectVoiceSession, startSession, widget, reconnect, and session errors.
@node-webrtc-rust — Open WebRTC SDK, signaling, and helpers — the native stack VoiceThere cloud runners and @voicethere/client use.

Browser and clients

Browser client — @voicethere/client — connectVoiceSession for local dev vs cloud, chat, embed widget, iframe mic/autoplay.
Browser CORS origins — Allow your production and staging sites to call the VoiceThere sessions API from the browser with a client key.
Embed widget designer — Dashboard Widget tab — presets, positions, customCss, positionOffset, fonts; Deploy publishes appearance JSON on the VoiceThere CDN (no client API key in public config).
Spoken chat stream — Stream agent speech into the embed chat bubble word by word (streamSpokenText) — dashboard toggle, widget config, and speakAndChat with stream: true.

Agents

Agent templates — Dashboard starter templates — blank, voice-starter, echo, voice-showcase, world-sync, world-sync-binary, game-sync, recording-consent, positional-tts, spatial-showcase — seed agent.js bundles and root package.json from @voicethere/agent for local npm scripts.
Data-only world sync — Sync positional data over a WebRTC data channel — JSON onDataChannelMessage, binary ArrayBuffer poses with onDataChannelBinary, and Redis-backed world snapshots.
Dashboard code editor — Edit source on the Code tab — Save, download saved source or compiled agent.js, then Compile & upload, Promote this build, and Deploy to cloud. Also covers package.json, Refresh types, public npm, and per-plan source and typings limits.
Agent environment & secrets — How AGENT_* environment variables and encrypted secrets are stored, named, and injected into your deployed voice agent.
Agent crash policy — disconnect_all vs restart_child: what happens when an agent child crashes, session isolation, and message buffering limits.

Voice

Advanced voice settings — Tune VAD, barge-in, TTS speed, STT lifecycle, and speech event delivery for cloud voice agents.
STT & TTS vendors — Speech-to-text and text-to-speech providers on VoiceThere cloud runners, including VoiceThere speech and third-party API vendors.
Speech latency benchmarks — Measured latency of VoiceThere speech and of ElevenLabs, Cartesia, Groq, OpenAI and Fish Audio: speech end to final, reply start, TTS first audio, with methodology and commands to reproduce.
VoiceThere voice — VoiceThere's own speech-to-text and text-to-speech on cloud runners, with no API keys. Usage is metered in VoiceThere credits. API id voicethere (local-sherpa is a deprecated alias).
Piper TTS pauses and pronunciation — Punctuation pauses and eSpeak [[phoneme]] overrides for VoiceThere Piper TTS — no SSML. Tune tts.speed and trailing silence in Advanced voice settings.
OpenAI speech on VoiceThere — Whisper and GPT transcribe models, Realtime live transcription, and OpenAI TTS on VoiceThere cloud runners with AGENT_OPENAI_API_KEY.
Deepgram speech on VoiceThere — Deepgram Listen STT over WebSocket and Aura or Flux TTS on VoiceThere cloud with AGENT_DEEPGRAM_API_KEY.
ElevenLabs speech on VoiceThere — ElevenLabs Scribe v2 Realtime STT and streaming TTS on VoiceThere cloud with AGENT_ELEVENLABS_API_KEY.
Cartesia TTS on VoiceThere — Cartesia Sonic TTS over WebSocket on VoiceThere cloud. Pair with a separate STT vendor. AGENT_CARTESIA_API_KEY.
Fish Audio speech on VoiceThere — Fish Audio Drama 3 TTS and transcribe-1 STT on VoiceThere cloud with AGENT_FISHAUDIO_API_KEY. STT sends finals only, no partials.
AssemblyAI STT on VoiceThere — AssemblyAI universal streaming STT on VoiceThere cloud. STT-only; pair with a TTS vendor. AGENT_ASSEMBLYAI_API_KEY.
Google Cloud speech on VoiceThere — Google Chirp streaming STT and Neural2 or Chirp3-HD TTS on VoiceThere cloud with service account or API key secrets.
Groq speech on VoiceThere — Groq Whisper file STT and Orpheus TTS on VoiceThere cloud with AGENT_GROQ_API_KEY. No live STT partials.
Azure Speech on VoiceThere — Azure conversation REST STT (finals only) and neural REST TTS on VoiceThere cloud with Azure Speech secrets.
AWS speech on VoiceThere — Amazon Transcribe streaming STT and Polly neural TTS on VoiceThere cloud with IAM access keys in project secrets.
Spoken language identification — Detect the caller’s spoken language with Whisper tiny and handle user_language events in your agent.
Switching language mid-conversation — Automatic and agent-driven language switches without dropping WebRTC, per-language models, wait messages while a new language loads, and example timelines.
Positional mix & orbiting TTS — Voice+Data mix groups, live client poses, and per-client TTS speaker poses — equal-power stereo with inverse-square distance.
Clip playback — Play HTTPS audio clips or small inline bytes from your agent — target one or all voice clients, poll play status, and mix with TTS and microphones.

Sessions and conversation

Session idle timeout — Auto-disconnect idle WebRTC peers, voice vs data-only timers, voice-activity and data-channel reset rules, and billing impact.
Conversation history — Store final user speech and agent TTS text for the dashboard Conversation tab — STT/TTS only, not DataChannel chat.
Conversation recording — Opt-in voice-session audio recording with dashboard playback, included minutes per plan, rollover bank or metered overage, and agent recording controls.

Project and API

Runner settings — Runner mode, warm pool, shared agent child, idle scale-down, and IPC debug — configure pool behavior and redeploy to apply.
Rollout modes — What a deploy does with calls in progress: drain lets them finish on the previous version, force ends them within 15 seconds of the new version being ready.
Project Redis — Private Redis for shared agent state across concurrent sessions — enable on Advanced & Ultimate, connect with ioredis via AGENT_REDIS_URL in onAgentStart.
Deployment history — See what happened during each deploy: when it was requested, when new calls moved to the new version, when the previous version finished its calls, and how it ended. In the dashboard, the API and the CLI.
Project metrics — Read the Metrics area of a project: session outcomes, time to connect, voice response time and finalize delay, agent errors, and the metrics API.
Dashboard guide — A tour of the dashboard: sidebar and project tabs, project header, the Get this project live checklist, the Settings tab, the session timeline with per-turn details, Metrics, Usage and the Cmd+K command palette.
Project usage — The project Usage page: sessions, billable time, error rate, credit use over time, voice usage per project, and how to buy credit packs.
Voice metrics — Voice metrics now live in the Metrics area of each project. This page points to the Voice section of the project metrics guide.
Inbound webhooks — POST HTTPS webhooks to your VoiceThere project — fan-out to every ready session worker, verify in your agent, Advanced & Ultimate plans.
Control plane API — Interactive OpenAPI reference for the VoiceThere REST API — projects, builds, deploy, promote, and rollback.

Account and billing

Billing & usage credits — How VoiceThere measures usage, subscription included credits, credit packs, burn order, and organization budgets.
Speech pricing (VoiceThere STT & TTS) — How VoiceThere speech is metered: STT audio minutes, TTS characters by voice quality, and per-voice catalog rates.
Terms of Service acceptance — How VoiceThere Terms of Service acceptance works for the dashboard, CLI, and API — including how long authentication cache can delay a refreshed acceptance.

Logs and errors

Agent logs & session errors — Structured agent logs and session errors in the VoiceThere dashboard and CLI — agentLog, persisted failures, search, soft rollover, and plan limits.
System log events — Log lines VoiceThere writes into your session agent logs for language checks, language switches and voice settings changes.
Session errors — Session errors are now browsed in Agent logs. Data-channel session_error events and onSessionError are unchanged.