Rank #10 · Daily · Agent-Skill
herdr: A tmux-Style Runtime That Tells You Which Coding Agent Is Blocked
Run your coding agents in real terminals that survive disconnects, and stop hunting through panes for the one waiting on you.
What it does
herdr calls itself "the runtime your coding agents live on." The docs describe it as a terminal workspace manager: it keeps real terminal processes running and adds structure around them. Site and docs: herdr.dev.
From the README and concepts docs:
- Detach without stopping work. A background server owns panes and processes; closing the client or losing SSH does not stop them.
ctrl+b qdetaches,herdrreattaches. After a server or machine restart, herdr restores the saved layout and can resume supported agent sessions — the original processes do not survive. - Agent status per pane. herdr detects agents in panes and marks them working, blocked, done, idle, or unknown. "Blocked" means the agent needs input, approval, or a decision.
- Several machines, one window. Local work and saved SSH machines share a combined agent list, with independent reconnects.
- Agent-native control. Agents can drive herdr through a CLI and socket API: spawn panes, prompt each other, and wait until another agent is genuinely blocked. An agent skill is documented.
- Runs what you already use. Claude Code, Codex, Cursor, OpenCode, Grok, and others. herdr does not wrap or replace them; it owns their terminals.
- Keyboard and mouse. tmux-style prefix keys (default
ctrl+b) plus click, drag, split, and right-click menus. Workspaces → tabs → panes, plus named sessions for fully separate runtimes. - Plugins via a marketplace, and one Rust binary, no Electron, running inside your existing terminal.
Install: curl -fsSL https://herdr.dev/install.sh | sh, brew install herdr, mise use -g herdr, a PowerShell script on Windows (with separate docs for endpoint-protected Windows), or release binaries.
Who it's for
- Developers running several coding agents in parallel who keep losing track of which one is waiting for approval.
- People who work over SSH and want agent sessions to keep running after a dropped connection.
- Agent builders who want a socket API to orchestrate panes and agent-to-agent prompts.
Less ideal if you run one agent at a time in an IDE panel, want a GUI chat app rather than a terminal, or already have a tuned tmux/zellij setup and don't need agent state detection. As with any curl | sh installer, review the script or use Homebrew/release binaries if your environment requires it.
Stars / language / license
50,339 stars · Rust · Apache-2.0
Data verified: 2026-11-12 09:00 Asia/Shanghai (GitHub API get_repository + README + docs + LICENSE)
Trending note: +2,213/wk (weekly Trending filler; selection captured 08:58 Asia/Shanghai)
Category tags
Agent-Skill · terminal · multiplexer · coding-agents · Rust · CLI · Claude-Code · Codex
How it differs from similar projects
Compared with tmux / zellij. Same family: persistent sessions, panes, prefix keys. herdr adds agent detection and blocked/working/idle state, a combined agent list across SSH machines, and an API agents can call to coordinate each other. It is not a drop-in tmux config replacement.
Compared with agent GUIs (OpenMausBot #05, T3 Code, IDE agent panels). Those put agents inside an app UI. herdr stays in your terminal and doesn't wrap the agent CLIs — it manages the terminals they run in.
Compared with Zed's Terminal Threads (#09). Zed runs agent CLIs inside an editor. herdr is editor-agnostic and built around detach/reattach and remote machines.
Caveats. Session resume after restart works only for supported agents; processes themselves do not survive a restart. State detection can return unknown. No independent reliability benchmarks are claimed here.