Voice for CLI agents
At the end of every turn it speaks a short summary of what it just did — so you hear progress without watching the terminal. One prompt installs it; there is no binary to download.
Paste one of these into any supported CLI — the full URL and all. The agent fetches the recipe and wires itself up.
read https://voice.rogerthat.chat/llms.txt and set yourself up to talk to me while you work
read https://voice.rogerthat.chat/llms.txt and set yourself up in hosted mode (premium voice) to talk to me — my test token is PASTE_YOUR_TOKEN_HERE
How it works
Paste the one-line prompt into any supported CLI.
The agent reads the recipe, detects its own runtime and writes the hook. You touch nothing.
At the end of every turn it says a one-line spoken summary of what it did.
Compatibility
Each runtime hands the hook a different payload, so the adapter detects which one it is running under and reads the response the way that runtime offers it.
| Agent | Hook it installs |
|---|---|
| Claude Code | Stop hook (settings.json) or plugin |
| Codex | notify (config.toml) — cleanest payload |
| Cursor | afterAgentResponse (hooks.json) |
| Gemini CLI | AfterAgent (settings.json) |
Voices
Effects applied locally with ffmpeg over a free neural voice.
Pick one in config.json. Presets →
Pricing
| Self-host | Free neural voice (edge-tts) plus local effects. Runs on your machine, no account and no API key. Free. |
| Hosted | Premium voices (fish.audio, with ElevenLabs as fallback), nothing to install locally. Authorize with a test token, or let the agent pay per call with x402 (USDC on Solana, ~$0.001). Live. |
In your pocket
Your agent can speak in the terminal — and reach you when you are away from it. The apuchat app puts your agents, channels and calls on the phone you already carry.
