Guards

A pixel-art monster in a side pane that eats tokens and shows context, limits and activity at a glance

Claude Clubhouse: a toolbar above the prompt, rooms for usage, agents, commands, tool rules and colors, and whole-window themes on macOS

An orange cat above the prompt that reacts to your commits, tests and builds

/live: full-duplex voice (Kyutai streaming STT with semantic end of turn, or Silero and Whisper; Kokoro TTS) through a sidecar; /vibe: director mode that delegates all work to subagents; /livevibe: talk with a fast voice front that hands the work to Claude as director.

An unofficial fan mod: a cozy pixel-art band above the prompt where Clawd acts out what Claude is doing, beside a scene of session meters

A desktop mascot overlay that reacts to turns, tool calls and subagents

While the AI works, a plain-language knowledge card above the prompt explains a concept from the current task. Click 我知道啦 for the next one.

Opt-in persistent task checklist: the agent asks to use it, the user approves, checks items off and alone can clear or disable it.

Clawd lives above your Claude Code prompt, among live cards for model, context, usage and project.

Claude Code buddy plugin: an ASCII companion above your prompt that remembers your rules and flags Claude's shortcuts

Retro games in a Claude Code pane that run while Claude works and freeze the moment it needs you.

Buddy is back: the retired /buddy terminal companion as a Claude Code mod, with new sprites and Haiku writing its lines.

Redacts secrets and PII from every row Claude Code stores, before the model reads it and before the transcript keeps it. Reach L0: no network, no processes, no files.

Explanations as visual pages in a pane beside the chat: diagrams drawn in place, the full page one click away. No model calls.

Asks before Claude edits a file another chat on this machine changed in the last 30 minutes. Each chat keeps its own small ledger; no model calls, no network.

Desktop heads-up display for the Pro plan: exact 5h/7d usage gauges, context gauge, turn receipts, usage in the spinner and compact tool cards. No estimates, no tokens spent.

Stops Claude reading a file over 1,000 lines whole and gives it an outline tool (definitions with line numbers) so it reads only the range it needs. A retry still reads it whole.

Stops Claude re-reading a file it already read in this conversation when the file has not changed since; a deliberate second attempt still goes through.

/receipt opens a pane with the exact tokens every turn of the session spent (uncached, cached, output, subagents) and the costliest turns. No tokens spent.

Moves tools you have not used lately behind ToolSearch, so their descriptions stop riding along on every request; core tools and anything you used in your last five sessions stay loaded.

Keep the last 40 Bash command strings and copy one for reuse.

Chart successful built-in edits grouped by file path.

Combine observed tool and turn counts with a manual handoff note.

Chart completed tool latency and failures by tool.