Corgi: a multi-agent terminal shows what each agent is doing, and F1 jumps to the agent with unread output

Corgi appeared on Hacker News on Oct 23: a terminal where [Codex](/tools/codex) and [Claude Code](/tools/claude-code) run from one window, with an activity feed for each agent. F1 switches to the agent with unread output, and spending plus CPU, memory and GPU load are tracked per workspace.