Corgi Agent Multiplexer Debuts on Hacker News

Corgi displays what agents are doing and provides a running feed of their activity, according to the post. It includes an F1 shortcut to move to the next unread agent. The tool also offers a unified launcher for Codex, Claude Code and Pi, with fast search and resume across sessions and inline usage tracking. The developer said these features were built around ergonomics he had wanted.

The terminal experience includes one-click copying of links, so users do not have to drag-select across multiple lines; Command-V pastes images even over SSH; and images get thumbnail previews. Corgi also tracks CPU, memory and GPU usage broken down by workspace. The post said the software assumes access to inexpensive inference services such as Luna and Haiku, and reuses the agents to organize things, including identifying an HTTP link, explaining a worktree or describing what each agent is doing.

In a personal note, the developer said he had been on a frustrating sick leave and created Corgi to challenge himself and see whether he could return to work. He said the answer was more mixed than he had hoped but that he was happy with the result. The developer said friends and he had used Corgi for a few weeks, that it still has rough edges and room to improve, and that feedback would be appreciated. The Hacker News entry had 4 points and no comments.

Editor's Summary

Corgi is a new Hacker News project that multiplexes AI coding agents and terminal sessions, adding activity feeds, quick navigation, unified account launchers and terminal conveniences. The developer says a small group has used it daily for weeks and is seeking feedback while acknowledging rough edges. It targets users with access to low-cost inference for organizing agent context.