Terse changes how Claude Code writes its answers: lead with the answer, use short lists, and avoid filler, slogans, and lengthy recaps. Its output style sets the rules; a UserPromptSubmit hook repeats a condensed reminder each turn so the style is less likely to drift during a long session. A separate, optional status-line meter shows context-window fill.

  • Install: In Claude Code, use /plugin marketplace add lowenbjer/claude-terse, then /plugin install terse@terse; it is also listed in Claude’s plugin directory. The meter requires a separate /terse:install-meter command and writes a statusLine setting.
  • Availability: MIT-licensed source, with no plugin fee listed. You still need access to Claude Code and its usual subscription or API billing. The author says the plugin itself makes no network calls; prompts still go to Anthropic through Claude Code.
  • Evidence, not a guarantee: The author’s benchmark reports shorter replies and lower list-price cost on a private-codebase prompt set. The public set ran each prompt once, and the newer 1.2 writing rules have no comparable public-set rerun. Do not assume those savings or answer quality transfer to your projects.

The October 2 release added plugin-directory packaging and privacy documentation, not new writing rules. This is specific to Claude Code, not a general Hermes or Codex plugin. HN discussion raises the reasonable concern that shorter replies might hide useful reasoning; the published comparisons measure style and cost, not independent correctness on real tasks.