Herdr: Agent multiplexer that lives in your terminal

That landed with people already running parallel agent workflows. The strongest use case was not opening a few extra local shells. It was keeping a persistent process on a machine, SSHing back in later, and recovering every active session from a laptop, phone, or tablet. Several people said that is where Herdr beats plain tmux in practice. It is easier to inspect and switch between agents, especially on touch devices, and it works well when some agents are local and others live in remote sandboxes the team already manages itself. The discussion also made clear why this category exists at all. For many users, one or two agents is still the limit, and tmux, Kitty, Zellij, or a couple of VS Code terminals are enough. The pressure shows up when agent latency is measured in minutes, not seconds, and developers start running five or more sessions round-robin across bugs, features, reviews, and deploy checks. At that point the problem shifts from pane management to awareness. You need to know which session is waiting for input, which one finished, and which one is stuck. Commenters kept coming back to Herdr’s notifications, status markers, persistent sessions, mouse scrolling, and workspaces as the features that make it feel meaningfully different from generic multiplexers. The mood was positive but pragmatic. People liked that Herdr stays close to the shell and existing infra instead of forcing a new cloud control plane. At the same time, plenty of commenters saw it as one option in a crowded field of tmux setups, editor-integrated tools, and web UIs. The dividing line was simple. If you already have a stable one-or-two-agent workflow, this solves a problem you may not have yet. If parallel agents are becoming operational overhead, Herdr looks less like bloat and more like missing plumbing.