claude-code - 💡(How to fix) Fix [A11y feature request] Built-in option to speak Claude's responses aloud

Summary

Add a built-in option to have Claude's responses spoken aloud in the Claude Code desktop app. This benefits blind users, low-vision users, and anyone working hands-busy or hands-free.

Proposal

Two patterns, either or both:

1. User setting (opt-in, off by default) — when enabled, completed Claude responses are read aloud automatically as soon as streaming finishes.
2. Keyboard shortcut — a binding (e.g. cmd+shift+R) that reads the most recent Claude response on demand.

Either pattern is well-established in accessibility tooling: think VoiceOver's "speak the focused item" or screen reader "read from here" commands.

Why built-in beats a hook workaround

A hook-based version (Stop hook + say) is theoretically possible but has problems:

• The Stop hook fires on /clear, /resume, /compact, etc., not just on actual response completion.
• The hook payload likely does not contain the response text, so it has to scrape the transcript file.
• Markdown and code blocks sound terrible when read by raw text-to-speech (literal asterisks, backticks, "hash hash hash" for headings) — would need clean message data and intelligent rendering.
• No clean way to interrupt when the user starts typing the next prompt.

A built-in implementation would have access to clean message data, settings UI, proper interrupt handling on user input, and nicer voices than say.

Context

I submitted the same feedback via the in-app /feedback channel and to [email protected] so the team can dedupe.

I am Lead Accessibility Architect at Paramount Streaming, blind, daily VoiceOver user. Happy to help spec, test, or pilot.