Multiplex AI agents in persistent terminal sessions

herdr describes itself as the runtime your coding agents live on. It is a single Rust binary, with no Electron, that runs inside whatever terminal you already use and owns the terminals of the coding agents you run. It does not wrap or replace those agents. It names Claude Code, Codex, Cursor, OpenCode, and Grok "and the rest" as examples of what it runs.

The main behaviors listed are:

• Detach without stopping work: herdr keeps terminals running in a background server when you close the client or lose your SSH connection. After a server or machine restart, it restores the saved layout and can resume supported agent sessions, but the original processes do not survive.
• Several machines, one window: local work and saved SSH machines sit together with a combined agent list and independent reconnects.
• Status at a glance: every pane is marked working, blocked, or idle, and when an agent stops and needs an answer, herdr says so, so you never hunt for the stuck one.
• Agent-native control: agents drive herdr through its command line and a socket API. They can spawn panes, prompt each other, and wait until another agent is genuinely blocked.
• Input styles: tmux-style prefix keys and mouse actions (click, drag, split) are both first-class, chosen per moment rather than per tool.
• Plugins: extend panes and workflows through a plugin marketplace.

Use it when you run several coding agents at once, work over SSH or across more than one machine, and need long-running sessions to survive a closed terminal or dropped connection. It is also a fit when you want one agent to be able to start and coordinate another through a CLI rather than by hand.

Be aware of its stated limit: after a server or machine restart the layout is restored and supported agent sessions can be resumed, but the original processes are gone, so it is not a way to keep a computation alive across a reboot. Sessions can only be resumed for supported agents, and the docs keep a supported agents page.

The input is your normal terminal workflow: you start herdr where the work lives, run your agents, and split panes. The output is a persistent multi-pane workspace with status marks per pane. The prefix key ctrl+b q detaches, and running herdr again reattaches. Installation on macOS or Linux is one line:

curl -fsSL https://herdr.dev/install.sh | sh

Alternatives listed are brew install herdr, mise use -g herdr, a PowerShell install script for Windows, and prebuilt binaries on the GitHub releases page. Then start it with the command herdr.

herdr works alongside the coding agents you already use and connects saved SSH machines into one window. Agents integrate through the CLI and socket API, and an agent skill is documented. Extensions arrive through plugins from its marketplace. Development uses cargo build --release, with just test for unit tests and just check for formatting, tests, and maintenance checks.

It is for developers who run several terminal-based coding agents in parallel, work on remote machines, and want their sessions to persist, with clear visibility of which agent is working, blocked, or idle. The project is licensed under Apache 2.0, and full documentation lives at herdr.dev.