Three tools rewrite Claude's output, one setting already does it

Three developers built tools that pipe Claude Code's replies through a second model before display, each within about two weeks of the others, per paddo.dev's writeup. Every one of those tools adds a second inference call and an external model dependency to each reply, a cost Anthropic's own settings already remove.

vomit rewrites Claude's display text through a local OpenAI-compatible model via a hook. claudish-to-english appends a plain-English rewrite generated by Ollama under each reply. Claudette routes the same rewrite through Gemini, using the Antigravity CLI.

Anthropic already ships a fix for this. The Concise output style, added in version 2.1.237, rewrites replies without a second model call, and error reports, security warnings, and destructive-action confirmations keep their full content in Concise mode. Set "outputStyle": "Concise" in ~/.claude/settings.json and start a new session to turn it on.

None of them needed a new model call. Each builder hit the same annoyance and missed the setting that already covers it, which points to a documentation gap, not a real limit in Claude Code's output style.