Heard

Your agents have a voice now.

Heard (heard.dev) is the voice layer for coding agents on macOS. It reads Claude Code and Codex progress aloud: when your agent finishes, fails or needs your approval, Heard tells you in a sentence or two, so you can step away and still know what's going on. The engine in this repo is open source (Apache 2.0) and runs with local Kokoro text-to-speech, no account needed. The Heard app adds cloud voices and cloud agents like Grok Bot, Meta Muse, Devin, Manus and ChatGPT, and on Power you talk back. Think Jarvis for your agents.

Comparing macOS coding-agent notification tools? A sound or a banner tells you something happened. Heard tells you what happened: a spoken summary of the work itself, across every session and every agent.

heard.dev  ·  Releases  ·  Discord  ·  Issues

What is a voice layer for Claude Code?

A voice layer lets your coding agent talk to you, not only you to it. Claude Code's /voice covers the input half. Heard covers the output half: it follows the session and tells you out loud what changed, what failed and what needs a decision. The Heard app adds input too (push to talk and hands-free dictation on Power), so the whole loop can run by voice.

Make Claude Code talk back, not just listen

Can you use text-to-speech with Claude Code? Yes. Claude Code's built-in /voice command is voice input: you talk and it types your prompt. Heard adds the other direction. Run heard install claude-code and Heard adds hooks (Stop, PreToolUse, PostToolUse, UserPromptSubmit) to ~/.claude/settings.json, then speaks a short summary of what the agent did instead of reading the whole reply back. Voices come from local Kokoro (free, on your Mac) or your own ElevenLabs or Speechify key. Hooks return immediately, so Claude Code never waits on speech.

Runs fully local

The engine in this repo is open source under Apache 2.0 and needs no account or API keys:

• Speech comes from Kokoro text-to-speech on your Mac (the model downloads once on first use).
• Without an Anthropic key, narration comes from local templates, so nothing about your session leaves the machine.
• It speaks; it does not listen. Push to talk and voice replies are app features on Power.
• Add your own keys only if you want ElevenLabs voices or LLM-written summaries.

git clone https://github.com/heardlabs/heard.git && cd heard
python3 -m venv .venv && source .venv/bin/activate && pip install -e .
heard install claude-code        # or: heard install codex

Full options in Self-host. Comparing local voice tools for coding agents? See open-source voice tools for coding agents.

See and hear it run

🔊 Turn sound on for demo

https://github.com/user-attachments/assets/d823a946-fb6f-438b-904f-aa66d4268ed1

Get the app

The app is the managed experience: download, sign in, and you're narrating. Heard runs the cloud voices and the narration brain for you, with no keys to manage. Every new account starts with a 14-day Power trial, no card required. Prefer to run it yourself with your own keys? See Self-host.

Have your coding tool install it (recommended)