DashVox brings voice control to coding agents on your devices

DashVox.ai has launched an iPhone app that lets users start and steer Claude Code and Codex sessions running on their own machine over SSH by voice. The app supports phone, CarPlay, Android Auto, Apple Watch, Wear OS, and a self-hosted server option, with Android beta coming soon.

DashVox brings voice control to coding agents on your devices

DashVox.ai has launched a voice-first app for controlling AI coding agents running on a user’s own machine. The company says the app can start and steer Claude Code and Codex sessions over SSH using spoken commands, and it is available now on the iPhone App Store.

According to the DashVox website, the app is designed for use on a phone, in a car through CarPlay or Android Auto, and on watches including Apple Watch and Wear OS. The company also says smart-glasses support, including Ray-Ban Meta, is being explored.

⚡ New to this?

This matters because it puts voice control on top of AI coding tools that usually expect a keyboard and screen. SSH means a secure way to connect to a computer remotely, so the app is not just chatting with an AI on the phone - it is controlling sessions running on another machine. For non-experts, the big idea is that coding assistants are moving from desktop apps into cars, watches, and other hands-free devices.

🦞 OpenClaw angle

If you run self-hosted agents, test whether your workflow can be controlled safely through voice and a mobile client instead of only a terminal. Keep the agent’s server local where possible, because DashVox’s setup shows users will expect data to stay on their own machine. If you build similar tools, separate “speak” and “approve” steps clearly so voice commands cannot trigger risky actions by accident.

The product is aimed at users who want to interact with coding agents without sitting at a keyboard. DashVox says voice is transcribed in real time, allowing users to ask questions, start coding sessions, or search the web. Responses are then spoken aloud through text-to-speech, while summaries can also be reviewed on the display.

The workflow is split into three steps, according to the company: open the app, speak the request, and listen to the response. DashVox says users can approve actions by voice as part of the same flow.

The company says the app works with multiple model providers, including Claude, GPT, and Gemini. That means it is not tied to a single large language model, even though the current examples on the site focus on Claude Code and Codex sessions.

DashVox also says it offers a self-hosted server option. The backend can run on a user’s own machine, and the company says all data stays local in that setup. The site describes the backend as cross-platform and built with Java 21 and Spring Boot, with H2 for self-hosting and PostgreSQL for the cloud version.

On the mobile side, DashVox says the iOS app is native SwiftUI and includes CarPlay support. For Android, the company says Android Auto is ready, and voice replies also work on Wear OS watches through connected-device notifications.

Authentication and cloud components are also part of the stack. The site lists Firebase Auth for cloud authentication, along with Google Cloud, Anthropic API, OpenAI API, and Gemini API.

DashVox says the iOS app is available now and that Android has been approved, with a public beta coming soon. The company is positioning the app as a way to control AI coding sessions from the phone, car, watch, and, eventually, smart glasses.

Source: HN Show HN ↗

More from OpenClaw News