Claude Code now has a screen reader mode

Fetching the source for concrete facts so the paragraphs stay accurate and invent nothing.Claude Code now ships a **screen reader mode**. The announcement from **ClaudeDevs** states that running `claude --ax-screen-reader` replaces the usual visual terminal UI with plain, linear text that screen readers can track. The same post names **VoiceOver** and **NVDA** as the kind of readers the mode is meant to work with. On Hacker News the item sat at **2 points** with **0 comments**, so the signal is early and thin on public discussion, but the product change itself is concrete: a dedicated accessibility path for the CLI, not a vague accessibility claim.

The mechanics are about presentation, not a second model or a separate app. Claude Code’s default experience is a rich terminal UI—status chrome, live updates, and layout that sighted users scan visually. Screen readers need a stable, sequential stream of text; rapid redraws and decorative UI often break speech output or dump noise like symbols and partial tokens. The new mode swaps that visual surface for linear text so the reader can walk the session top to bottom. Enabling it is opt-in via the **`--ax-screen-reader`** flag, which keeps the default TUI for people who want it and gives a plain path for people who need it.

For engineers who build or depend on agentic coding tools, this is a usability gate, not a nice-to-have. Blind and low-vision developers use the same CLI workflows—open a repo, ask for edits, approve tools, review diffs—as everyone else. If the interface cannot be spoken reliably, those workflows stop at the UI layer even when the model and tools work. Teams that standardize on Claude Code for pair-programming-style automation now have an official way to keep that path open for coworkers who run VoiceOver or NVDA. Accessibility here is product surface area: who can complete a session without restarting the terminal or the screen reader.

Competitive context is the terminal-first coding-agent market. Tools in the Claude Code / Codex class sell speed and depth of repo work, but most of that value assumes a visual TUI. Gemini CLI and similar agents have already been pressed for **screen-reader** flags for the same reason: streaming tokens, spinners, and ANSI styling fight assistive tech. Claude Code matching that pattern with an explicit **`--ax-screen-reader`** switch moves it into the set of agents that treat accessibility as a first-class CLI mode rather than an afterthought. Market pressure is straightforward: if one major agent is usable under NVDA or VoiceOver and another is not, hiring and tool choice for mixed-ability teams tilt accordingly.

Practical takeaway: if you use Claude Code and either rely on a screen reader yourself or support people who do, try `claude --ax-screen-reader` in a real session—tool prompts, long streams, multi-file edits—and confirm speech stays stable. If you maintain internal docs or onboarding for the tool, document the flag next to install and auth steps so it is discoverable. Watch next for how complete the plain-text surface is under real assistive setups (permission prompts, status, errors) and whether the mode stays a single flag or grows into broader a11y work across the Claude Code CLI.