VOXHELM Join the beta
Private beta · iOS, Android, macOS

Talk to your AI agent. Out loud.

Voxhelm is a voice app for AI agents. Create an agent, give it a face and a voice, then call it and just talk. It answers in real time, with a live transcript of everything said.

Join the beta See how it works

Native apps for iPhone, Android and Mac. Agents are hosted for you.

Live00:42
An illustrated avatar of a young woman with long auburn hair
Ava
Personal assistant

YouWhat's on for tomorrow?

AvaDentist at nine, then you're clear until the team call at two.

How it works

From sign-up to talking in a couple of minutes

No servers to run and no keys to wire up. Voxhelm hosts the agent, the speech and the call.

  1. Create an agent

    Give it a name, pick a template for what it should do, then choose a face and a voice. Voxhelm starts it in its own container.

  2. Call it

    Tap call and talk naturally. Your speech is transcribed as you go, the agent replies, and you hear the answer in its voice.

  3. Keep the conversation

    Every call and chat is kept as a transcript on your device. Prefer typing? Chat with the same agent instead.

Features

What works today

Everything below is in the beta build now. Things still on the way are on the roadmap.

Real-time voice

Streaming speech recognition and speech synthesis over persistent connections, so the agent starts speaking before its full answer is written.

Live transcript

See what you said and what the agent said as the call happens, so you don't miss a detail.

Faces that talk

Pick from a range of avatars, from people to robots to a cheerful crustacean. Their lips move in time with what the agent says.

Hosted for you

Each agent runs on OpenClaw in its own container, set up from a template. Nothing to install or keep running yourself.

Chat when you can't talk

Message the same agent by text instead. Handy in a meeting or on the train.

History on your device

Call and chat history is stored in the app on your phone or Mac, and you can delete it at any time.

Avatars

Pick a face for your agent

A few of the avatars in the app today. Generating your own is on the roadmap.

Young Woman
Photo Man
Older Woman
Clawford
Nimbus
Mossling
Chrome Bot

Roadmap

Where Voxhelm is heading

The beta keeps things small on purpose. Here's what's next, roughly in order. Beta testers help decide what comes first.

NowIn the beta

  • Hosted agentsCreated from templates, one container each
  • Voice calls and chatReal-time, with live transcripts
  • Talking avatarsLip-synced to the agent's voice
  • iOS, Android and macOSTest builds for invited beta users

NextBeing built

  • Bring your own OpenClawLink the agent you already run and talk to it by voice
  • More voicesA wider choice of voices and accents
  • Agent editorChange an agent's instructions, skills and settings from the app
  • Open sourceThe core services published under the MIT licence

LaterOn the horizon

  • Your own avatarGenerate a face from a description or a photo
  • Hermes agentsVoice for agent frameworks beyond OpenClaw
  • Phone callsCall your agent from any phone, or let it answer for you
  • CarPlayHands-free calls to your agents while you drive

FAQ

Questions

What does it cost?

The beta is free. Voice minutes may be capped per person during the beta to keep costs predictable. Paid plans will come after the beta, and testers will hear about them first.

What is OpenClaw?

OpenClaw is an open-source framework for running a personal AI agent. Voxhelm runs a hosted OpenClaw agent for you and adds the voice, the app and the avatar on top.

I already run my own OpenClaw agent. Can I use it?

Not yet, but it's the next big feature. You'll link your existing agent in the app and talk to it by voice, with no migration. If that's why you're here, say so when you join the beta.

Which devices are supported?

iPhone, Android phones and Macs. Beta builds go out through TestFlight on iOS, Google Play internal testing on Android, and a signed download for macOS.

Where is my data processed?

Voxhelm's services and your hosted agent run on AWS in Sydney, Australia. Speech recognition and speech synthesis are handled by Deepgram, and your agent's replies come from the AI model it's configured to use. See the privacy notice for details.

How do I get in?

Join the beta below. Invites go out in small waves so each group gets proper attention.

Join the private beta

Invites go out in small waves to people who want to talk to an AI agent every day, especially if you already run OpenClaw.

Beta sign-ups open soon

Sign-ups open shortly. Check back here.