RFC: Extensible Companion Framework — let the community build what /buddy started

Problem

/buddy proved something important: developers form emotional connections with their tools. The 40+ issues filed in 48 hours after its removal aren't about a penguin — they're about a CLI that felt alive.

But bringing back a hardcoded Easter egg isn't the right solution. The right solution is bigger.

Proposal: Companion Framework

A lightweight, extensible system that lets Claude Code render ambient TUI companions — first-party or community-built.

// ~/.claude/settings.json
{
  "companion": {
    "enabled": true,
    "species": "penguin",     // built-in: penguin, yak, crab
    "custom": "~/.claude/companions/my-buddy.json"  // community
  }
}

Spec (minimal):

• Companion = ASCII sprite + idle animations + optional context reactions (e.g., wobble on test failure, sleep on long build)
• Zero token cost — purely TUI-rendered, never injected into context
• Hook integration — companions react to existing hook events
• claude companion list / claude companion install <name>

Why this is different from "bring back /buddy"

This isn't nostalgia. This is a platform feature:

• Community creates companions → ecosystem engagement
• Companions react to hooks → hooks adoption increases
• Customization → users invest identity into their setup → retention
• Extensible → Anthropic ships once, community maintains forever

Why Anthropic should care

The terminal is the last unbranded touchpoint. VS Code has themes. GitHub has contributions graphs. Slack has custom emoji. Claude Code has... a blinking cursor.

Give us the penguin back. But this time, let us build the zoo.