herdr
tmux rebuilt for the era of several coding agents at once — every pane marked working, blocked or idle, and a socket API the agents themselves can drive.
Run four agents at once and the bottleneck stops being the agents. It becomes you, cycling through panes to find which one stopped three minutes ago waiting for a yes.
tmux can’t help: it was built before any of this existed and has no concept of what’s happening inside a pane. herdr is the same idea with that concept added — a single Rust binary, Apache-2.0, running in whatever terminal you already use.
Install
curl -fsSL https://herdr.dev/install.sh | sh
herdr
Also brew install herdr or mise use -g herdr, which are the better options if you’d
rather not pipe a remote script into a shell. Windows gets a PowerShell installer, and
there are plain binaries on the releases page.
Start it where the work lives, run your agents, split panes, walk away. ctrl+b q
detaches; herdr reattaches. If tmux is in your fingers, so is this — and the mouse
works too, click, drag and split, which tmux users spend years pretending they don’t
want.
Never hunt for the stuck one
Every pane carries a state: working, blocked, or idle. When an agent stops and needs an answer, the sidebar says so rather than leaving you to notice. That single feature is most of the value — supervising parallel agents is an interrupt-driven job, and everything else is a way of not missing the interrupt.
It runs what you already run — Claude Code, Codex, Cursor, opencode, Grok and the rest — and importantly doesn’t wrap or replace them. It owns their terminals. Nothing about your agent setup changes.
What “survives a disconnect” actually means
A background server holds the terminals, so closing the client or losing SSH doesn’t stop the work. Worth being precise about the limit, because summaries of this tool routinely overstate it:
After a server or machine restart, herdr restores the saved layout and can resume supported agent sessions — but the original processes do not survive.
So: drop your connection, close the lid, reattach from another machine — fine, the work kept running. Reboot the box — you get your layout back and supported agents resume, not the same processes carrying on. Plan long jobs accordingly.
Several machines can also live in one window: local work alongside saved SSH machines, one combined agent list, reconnecting independently.
The part that isn’t just a nicer tmux
herdr exposes a CLI and a JSON socket API that agents drive themselves. An agent can spawn a pane, start another agent in it, prompt it, and wait until that one is genuinely blocked rather than polling for output that may never come.
That turns the multiplexer into coordination infrastructure. The interesting version isn’t you watching four agents — it’s one agent fanning out work to three others and being told, by the runtime, exactly when each needs something. There’s a plugin marketplace on top of that for extending panes and workflows.
Against the other two here
| Best at | |
|---|---|
| Luvus | A terminal mission control that runs, watches and resumes agents for you |
| Orca | A desktop app giving each task its own worktree, terminal, browser and diff |
| herdr | The layer underneath: it owns the terminals, and the agents can program it |
Orca is the one to pick if you want isolation per task and a GUI to review diffs in. herdr is the one to pick if your work already lives in a terminal over SSH and you want the agents themselves to be able to coordinate.
It hit #1 on GitHub Trending in June 2026 and raised a seed round in September — fast enough that the version you install will have moved past this page. Check the docs for anything that matters.
Next
Whichever agents you run, they work better with a map of the repository → Graft
Last updated 20 Sep 2026, 00:00 UTC.