CodeTalk — Development skill for Claude Code
A spoken reflection layer for Claude Code CLI — gives Claude a voice using edge-tts.
How to install CodeTalk
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open ronle/CodeTalk and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What CodeTalk does
A spoken reflection layer for Claude Code CLI — gives Claude a voice using edge-tts.
Alternatives in Development
- CORS — Scan an endpoint for CORS misconfiguration — arbitrary-origin reflection, null-origin trust, credential exposu 4.5k ★
- Arreflect — AgentRecall consolidation & reflection — periodic triage of recurring corrections; proposes rule changes, neve 370 ★
- Jianing Skills — Agent skills for Claude Code and Codex: self-reflection, content, slides, WeChat formatting 328 ★
README
CodeTalk
A spoken reflection layer for [Claude Code](https://docs.anthropic.com/en/docs/claude-code) CLI. CodeTalk gives Claude a voice — it speaks brief observations, task announcements, and non-obvious insights aloud while you work.
How It Works
Claude embeds short spoken lines directly in its responses using a distinct format. A Stop hook extracts the text and plays it through your speakers via [edge-tts](https://github.com/rany2/edge-tts) (Microsoft's free neural TTS).
Your normal response text here...
---
> *This is what gets spoken aloud — a brief observation the model decided was worth saying.*
No second LLM call. No API key. The main model decides when to speak and what to say.
Voice Behaviors
| Type | When | Example |
|---|---|---|
| Task start | Beginning a non-trivial task | "Refactoring the scanner to use connection pooling." |
| Task complete | Finishing a task or milestone | "Migration done, both tables updated." |
| Reflection | ~30% of substantive responses | "Removing that API key also eliminates the Sonnet charge risk." |
Silence is the default. Not every response gets a spoken line.
Setup
1. Install edge-tts
pip install edge-tts
2. Add the hook to Claude Code
Add to `~/.claude/settings.json`:
{
"hooks": {
"Stop": [
{
"matcher": ".*",
"hooks": [
{
"type": "command",
"command": "python \"/path/to/CodeTalk/codetalk/hooks/codetalk.py\"",
"timeout": 30000
}
]
}
]
}
}
3. Add voice instructions to your project
Copy the `## Voice Instructions (ACTIVE)` section from this project's `CLAUDE.md` into your own project's `CLAUDE.md`. This tells Claude when and how to embed spoken lines.
Configuration
All settings are optional environment variables:
| Variable | Default | Description |
|---|---|---|
CODETALK_VOICE |
`en-US-Andr |
Related Skills
Terminal Talk
Hands-free voice output for Claude Code on Windows. Free, MIT licensed, no signup, no accounts. Microsoft Edge
Claude Voice System
Add self-learning voice notifications for Claude Code with edge-tts, phrase pools, priority queues, and anti-r
Odin Reflector
Codex CLI meta-reflection layer for Outline-Driven Development. Add-on for the Claude plugin.
Voizecode
Talk to Claude Code by voice — laptop CLI + Deno relay + Next.js client, streaming TTS, multi-session, barge-i
PDF To Tts Zh Skill
Turn a PDF into a single narrated Traditional Chinese audiobook MP3 via edge-tts. Install: npx skills add htli
Vibe Tts
独立 TTS(配音)技能:edge 预置音色与 IndexTTS-2.5 本地声音克隆,文本加风格一步产出 mp3
Related Agents
Voice Critic
Reads a Written as Spoken draft cold, as a tired Egyptian marketer scrolling at 11 pm, and returns line-level
Voice Only
Minimal voice-only example — converse only; all other tools (Bash, Read, Write, Edit, …) are disabled by desig
Voice Engineer
GAIA voice interaction specialist. Use PROACTIVELY for Whisper ASR, Kokoro TTS, the Talk SDK, speech-to-speech