Skip to main content

What is Voxa

Voxa is a voice operator for Claude Code. You call your laptop from your phone, tell it what you want done in plain language, and hang up. Claude Code does the actual work on your laptop while you walk away from your desk, and when it finishes (or needs a decision from you), your phone rings.

Under the hood, Voxa isn't the thing writing your code. It's a voice layer that sits between you and Claude Code: a Gemini Live voice model listens to you, relays your instructions to a real Claude Code session, and speaks back the result. It can attach to a terminal you already have open, start a fresh session in a project folder, switch you between several running sessions, and hold anything destructive (a git push, for example) for your explicit approval before it runs.

The mental model​

Three parts, one loop:

   phone                laptop                  Claude Code
  -------              --------                -------------
  you talk   --voice-->  Voxa server  --task-->  does the work
                             |                        |
  phone rings <---calls back-+<------ finished --------

You speak to the phone. The phone's audio goes to the Voxa server running on your laptop. The server hands your request to Claude Code and watches it work. When Claude Code is done, the server calls your phone back and tells you what happened.

What's supported today​

  • macOS is the only supported laptop platform right now. Windows and Linux are on the roadmap.
  • The native iPhone app gives you agent-initiated calls (your phone actually rings, like a phone call) and the smoothest pairing experience.
  • Any phone browser works too. Voxa's laptop server serves a plain web client, so you can open a pairing link on Android or anything else with a microphone and a browser, no app install required. Native Android support is on the roadmap.

Open source scope​

The CLI, the laptop server, and the phone web client are MIT-licensed and live in this repo. You can read every line of what talks to Claude Code, what goes to Google's Gemini Live API, and what (if anything) leaves your laptop. You can self-host the whole thing with your own Gemini API key, no Voxa account required.

The native iOS app, and the hosted relay/push/billing service that back the zero-config experience at voxa.space, are proprietary. That's the part that lets you install voxa, get a phone call, and never touch an API key: it's a convenience layer on top of the same open core.

Where to go next​

  • Quickstart: install Voxa and make your first call.
  • How Voxa works: the components, the topologies, and the life of a voice session.
  • Security: what data goes where, and what never leaves your laptop.
  • Self-hosting: run Voxa with your own Gemini key instead of the hosted relay.
  • Reference: every environment variable and install path, documented.