Anthropic: Claude Code v2.1.208 and v2.1.209 — screen reader, vim remaps, and wrapper support

Screen reader mode: accessibility as an opt-in setting

Claude Code v2.1.208 introduces screen reader mode — an opt-in mode that switches the interface to a plain-text display without visual elements that interfere with speech synthesizers. A screen reader is an assistive technology used by blind and visually impaired users to read screen content. Activation is possible in three ways: the --ax-screen-reader flag at startup, the CLAUDE_AX_SCREEN_READER=1 environment variable, or the "axScreenReader": true setting in the user settings file. Unlike earlier versions that had no dedicated support for assistive technologies, this version explicitly addresses that user group for the first time.

Vim insert-mode remaps and corporate process wrapper

The same version also introduces vimInsertModeRemaps — a new settings option that enables mapping two-key sequences within vim insert mode. Example: the sequence jj can be mapped to Escape, which is standard practice among vim users who want to exit insert mode without reaching for the Escape key. This type of remapping was not previously supported within Claude Code’s vim mode.

Wrapper executable for every self-spawned process

The third new feature in version 208 is the CLAUDE_CODE_PROCESS_WRAPPER environment variable. Corporate IT teams that use their own launchers — for example for logging, security checks, or proxy configuration — can now ensure that every self-spawned process (agent view, background service) passes through that wrapper executable. In earlier versions the wrapper only covered the initial call, while self-spawned processes bypassed corporate launcher policies.

Hotfix v2.1.209: reverting an overly broad guard

Just a few hours after 208, Anthropic released hotfix v2.1.209 to fix a critical bug in claude agents background sessions. Dialog windows such as /model were completely blocked — users could not change the model or access other dialogs from background sessions. The cause was an overly broad guard introduced in version 208; the hotfix reverts it and restores previous behavior. The speed of response — a hotfix within the same day — follows the pattern Anthropic applies to regressions that block standard workflows.