Read Claude Code's replies aloud, and see which sentence is being spoken, highlighted in place in the reply. Built as a Claude Code mod (a plugin with a hooks module), terminal first. The interface is in Spanish.
At the prompt of a Claude Code terminal session:
/plugin install read-along --marketplace nachodd/read-along
From a shell:
claude plugin install read-along --marketplace nachodd/read-along
Requires Python 3.9+ on your PATH as python3. No packages: the engine is the standard library.
Out of the box it uses your system voices (say on macOS). For a much better local neural voice, install Kokoro once (~340 MB, no account, 54 voices in 9 languages):
python3 <plugin folder>/engine/speak.py --setup kokoro
Cloud voices (Speechify, ElevenLabs, OpenAI) work too with your own API key; see the engine's config below.
/read-along |
read the last reply |
/read-along 3 |
read the third reply counting back from the last (like /copy 3) |
/read-along <text> |
read that text |
/read-along-stop |
stop |
/read-along-lang [auto|es|en] |
voice language; auto picks Spanish or English per reply |
/read-along-voice [es|en] [voice] |
list voices, or pick the one for a language |
/read-along-speed [0.5–2] |
speed |
/read-along-auto [on|off] |
read every reply when its turn ends |
/read-along-status |
provider, language, voices, speed |
On screen:
- Right of the prompt footer:
🔊 read/■ stopandlang: auto|ES|ENbuttons. - While reading: the reply keeps its formatting; the sentence being spoken is drawn as inline code (in color), or marked with
▸when it holds markup of its own. A band above the prompt shows progress, language, voice and speed, with Parar and − / + speed buttons; a new speed is heard from the next sentence, no restart. - Hovering a reply shows a 🔊 button that reads just that block.
Clicking needs the fullscreen terminal (mouse support); every action also has its command.
engine/speak.pyis the TTS engine from claude-read-aloud by Michael Gifford (MIT), vendored unchanged: providers, Kokoro setup, gapless chunked playback.engine/read_sentences.pysplits a reply into sentences, speaks them in order (synthesising the next while the current one plays) and prints which one is playing.hooks/register.tsxis the mod: commands, the buttons, and theAssistantMessagerender hook that marks the current sentence in the reply's own markdown.
Voices, provider and speed default from speak.py's config at ~/.config/claude-read-aloud/config.json; the mod's own choices (language, voice per language, speed, auto-read) are kept per Claude Code profile.
MIT. engine/speak.py is MIT, © Michael Gifford, see engine/LICENSE-claude-read-aloud.