noisy-coding
Official# Noisy Studio
Talk to your coding agent while it works. Noisy Studio is a native desktop app
with speech recognition, spoken replies, per-conversation voice and character
settings, and a dashboard for the conversations you are following.
https://github.com/user-attachments/assets/331ca147-b325-4b89-9704-a3bba206f2f8
## Install
Download the macOS Apple Silicon app from [Releases](https://github.com/noisy/noisy-studio/releases),
move **Noisy Studio.app** to Applications, and open it. Complete provider and
audio setup in the app. The app includes its own Python engine; ordinary Claude
installation needs neither Python nor a separate service manager.
For Claude Code, install the companion plugin:
```sh
claude plugin marketplace add noisy/noisy-studio
claude plugin install noisy-studio@noisy
```
Restart Claude Code and ask it to set up Noisy Studio voice. Keep the app and
plugin versions together: the plugin launches the engine bundled with the app.
The internal plugin name remains `noisy-studio` for compatibility.
[Installation and troubleshooting](docs/INSTALL.md) covers permissions,
providers, updates, and a spoken round trip. The [Codex preview](docs/codex.md)
uses the same app and currently requires `uv` for its integration processes.
macOS is the validated release platform; other platforms are source development.
## Development
Use [local development](docs/local-development.md) for the isolated daemon on
7765 and dashboard hot reload. The installed app owns 9765 and has separate
settings. See [desktop packaging](docs/desktop-app.md) for signed app builds.
- [Agent hooks](docs/hooks.md)
- [Ports](docs/ports.md)
- [Product naming](docs/rebranding.md)
- [Storybook](https://noisystudio.ai/storybook/) - how the dashboard and the
widget are meant to look, published from `main`. Run it locally with
`cd dashboard && npm run storybook`. Its title shows the commit it was
built from, so a story can be cited by URL and version.
API credentials belong in the app's provider settings, never in chat or source.
Offline providers download their models on first use and then run locally.
## Upgrading to the V3 technical names
V3 uses `noisy-studio` throughout packages, commands and integrations. Older installations need an explicit upgrade; see the [agent-friendly upgrade and rollback guide](docs/rebranding.md) before replacing an existing installation.
TDQS
Scored across 4 tools
The set is mostly distinct: speak delivers a blocking utterance, announce is fire-and-forget, change_voice and list_voices have clearly separate roles. speak and announce both produce speech, so their overlap could cause an agent to misselect when the blocking behavior matters, but the descriptions offer strong guidance.
Naming is simple and readable with all verbs as commands, but it mixes one-word verb names (speak, announce) with verb_noun patterns (change_voice, list_voices). This minor inconsistency is not confusing and the style remains predictable.
Four tools is well-scoped for a voice/speech server: each tool covers a necessary function—speaking, quick updates, voice switching, and voice enumeration. There is no bloat or obvious missing core capability for the stated purpose.
The tool set covers the main lifecycle of spoken interaction: speak, gently tell what you're doing, switch voices persistently, and discover available voices. One possible gap is lack of a way to query the currently active voice, but this is a minor issue that does not block common workflows.