wtf is herdr?

Herdr is a terminal application for people who run multiple AI coding agents at once, like several instances of Claude Code working on different tasks. It shows all of your agents in one place and tells you at a glance which ones are blocked waiting for input, which are still working, and which have finished. Instead of showing a simplified summary, it displays the real terminal output of each agent, so you see exactly what each one is doing. One of its main features is that agent sessions keep running even after you detach from the terminal. You can close your laptop or disconnect over ssh, come back later, and reattach to find your agents still working, with their state intact. It supports both keyboard shortcuts similar to the tool tmux and normal mouse actions like clicking and dragging, so people can use whichever fits the moment. Herdr also exposes a socket based API that the agents themselves can use, letting one agent spawn new terminal panes, read output from other panes, or wait for another agent to finish before continuing. This means agents can coordinate with each other through Herdr rather than only being watched by a human. The tool also supports plugins that extend how panes and workflows behave. It is written in Rust and ships as a single binary with no Electron or other heavy runtime, so it runs in whatever terminal you already use on Mac, Linux, or a Windows beta. Installation is a single command line script, or available through Homebrew and other package managers. The project is dual licensed: free under the AGPL license, with a separate commercial license available for organizations that cannot use AGPL terms.