DashVox – Voice Control AI Coding Agents From Your Car

TL;DR

TL;DR: DashVox is a voice-first app that lets you start, steer, and monitor Claude Code and Codex sessions remotely — from your iPhone, CarPlay dashboard, Apple Watch, or Android Auto.

Source and Accuracy Notes

⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.

• Project page: dashvox.ai
• HN launch thread: news.ycombinator.com/item?id=48356768
• Self-hosted docs: dashvox.ai/self-hosted
• License: not publicly declared — binary-only distribution, no open-source repo detected
• Source last checked: 2026-07-14 (website content verified via curl)

What Is DashVox?

DashVox bridges voice interfaces — CarPlay, Android Auto, Apple Watch, Wear OS — to AI coding agents running on your own machine over SSH.

The core proposition is hands-free coding agent control. You speak a command on your phone or through your car’s dashboard, and DashVox translates it into actions on Claude Code or Codex running on a remote dev server. Responses come back as text-to-speech, so your eyes stay on the road.

The product targets developers who want to manage long-running coding tasks (builds, tests, migrations) from contexts where typing is impractical — commuting, driving, or moving around without a laptop. The agent can autonomously spin up multiple sessions (frontend, backend, tests) and coordinate between them based on voice instructions.

Setup Workflow

Step 1: Install Java 21

The DashVox server is a standalone Java JAR. Any Java 21 JRE works.

# macOS
brew install openjdk@21

# Ubuntu/Debian
sudo apt install openjdk-21-jdk

Step 2: Download the server JAR

curl -LO https://dashvox.ai/downloads/dashvox-server.jar

Or use the platform-specific installer scripts:

# macOS / Linux
curl -LO https://dashvox.ai/downloads/start.sh && chmod +x start.sh && ./start.sh

# Windows
# Download start.bat from https://dashvox.ai/downloads/start.bat

The current latest JAR version is v1.0.95 (verified from the self-hosted downloads page).

Step 3: Run the server

java -jar dashvox-server.jar

The server starts on port 8081 by default. On first run it shows the LAN address your phone app should connect to.

Requirements (from official self-hosted page):

| Resource | Minimum | |---|---| | Java | 21 or later | | RAM | 512 MB | | Disk | 100 MB for the JAR | | OS | Windows, macOS, Linux |

Step 4: Connect the mobile app

1. Download DashVox for iOS from the App Store (Android coming soon)
2. Point the app at your server’s LAN address (or port-forward for remote access)
3. Pair your SSH credentials for the dev machine running Claude Code or Codex

Step 5: Configure AI agents

The self-hosted backend ships with an embedded H2 database by default. For the cloud tier, PostgreSQL is used.

On the server side, DashVox communicates with Claude Code and Codex over SSH. It manages session lifecycle (start, stop, approve prompts, check output) entirely by voice command.

Platform Support

| Platform | Status | |---|---| | iOS | App Store — download | | Android | Coming soon | | CarPlay | Available on iOS | | Android Auto | Coming soon (approved) | | Apple Watch | Via mirrored iOS notifications | | Wear OS | Coming soon |

Key Features

Voice-First AI Chat — Talk to any LLM (Claude, GPT, Gemini) via the phone app. Responses are read aloud via TTS. No screen required.

Remote Session Management — SSH into your dev machine and manage multiple AI coding sessions simultaneously. Start, stop, and switch between sessions by voice.

Agent Mode — Tell DashVox to spin up separate frontend, backend, and test sessions. The AI coordinates them autonomously based on your instructions.

Web Search Integration — The AI can search the web on your behalf and read back summaries — useful for checking docs while your hands are occupied.

Self-Hosted Server — The backend runs locally with no Firebase dependency in standalone mode. No account required. All data stays on your machine. Cloud mode adds Firebase Auth and stores SSH keys server-side.

Security Notes

The self-hosted build (standalone mode) does not require or use Firebase. Your conversations, SSH keys, and session data stay on your own hardware.

The cloud tier uses Firebase Auth for sign-in and stores SSH keys server-side to connect on your behalf. At-rest encryption for cloud stored data is listed as planned, not yet implemented.

For production self-hosted use, consider running on an isolated network segment or VPN, since the server listens on LAN by default.

FAQ

Q: Is the source code open source? A: No. DashVox is distributed as a binary JAR only. There is no public GitHub repository. The license is not declared on the website.

Q: Can I use this without CarPlay or a smart vehicle? A: Yes. The iOS phone app works independently with full voice control. CarPlay and Android Auto are extra convenience layers, not requirements.

Q: Does DashVox work with coding agents other than Claude Code and Codex? A: The website specifically mentions Claude Code and Codex. Whether it works with other SSH-based coding agents depends on the CLI interface they expose — not officially documented.

Q: How does the self-hosted server communicate with the phone? A: The phone app connects to the self-hosted server over HTTP/WebSocket on the LAN address you configure. For remote access outside your network, port forwarding or a VPN is needed (not built-in).

Q: What happens to my data in cloud mode? A: In cloud mode, Firebase Auth handles sign-in and SSH keys are stored server-side. Conversational data may be processed by the AI providers (Anthropic, OpenAI, Google) per their respective privacy policies.

Conclusion

DashVox solves a real ergonomic problem: managing long-running AI coding agents when you cannot be at a keyboard. CarPlay and watch integration make it genuinely hands-free — useful for developers who spend time commuting or moving around.

The self-hosted tier is compelling — Java 21 JAR, no account, H2 embedded by default, all data local. If you want voice control of your Claude Code sessions without surrendering anything to the cloud, the setup is straightforward.

The trade-off is the closed-source, binary-only distribution. There is no way to audit the server code, and the lack of a declared license means the usual open-source freedoms (reuse, modification, forking) do not apply.

If that is acceptable, DashVox fills a genuine niche for hands-free AI coding agent control across multiple device contexts.