An open source tool that lives in your terminal and bundles multiple coding agents, such as Claude Code and Codex, into a single screen. It runs as a single binary written in Rust, with no Electron, no account signup, and no telemetry. Each agent runs in its own real terminal, and a background server holds the sessions, so work keeps running even when you close your laptop or lose your network connection. You can reattach later over SSH or from your phone and pick up right where things stood. A sidebar shows the state of every agent ─ working, blocked, or done ─ so you no longer have to cycle through panes hunting for the one that stopped. Since its release in March 2026 it has reached the top of GitHub Trending and has collected well over 10,000 stars.
Key Features
- Sessions that stay alive: A background server owns the agents’ terminals. Close the terminal app, drop the network, or restart the machine, and your agents either keep running or resume as they were
- A sidebar that shows state at a glance: See in one list whether each agent is working, blocked waiting for input, or finished. No more opening panes one by one to check
- Runs the agent CLIs you already use: Claude Code, Codex, Cursor, opencode, Grok, and others keep working exactly as before. Herdr only manages the terminals; it does not wrap or alter the agents themselves
- Extensible through a Socket API and plugins: The CLI and the local socket API are unified, so agents can split panes or start other agents themselves. Plugins are supported for adding further capabilities
- Reattach over SSH or from your phone: Run it on your laptop, an always-on mini PC, or a rented box ─ wherever your code and terminals live ─ and reconnect from wherever you happen to be
- A dependency-free single binary: Written in Rust and available for macOS, Linux, and Windows. Install it with the install script, Homebrew, or mise. It runs inside the terminal app you already use (iTerm2, Kitty, Alacritty, WezTerm, Ghostty, and others)
Pricing
| Plan | Price | Highlights |
|---|---|---|
| Open source | Free | Full functionality. Source code published on GitHub |
| Commercial license | Contact required | Offered as an option for organizations that cannot comply with the open source license terms |
Pricing is accurate as of August 2026. License wording differs between sources, so check the official site and the LICENSE file in the GitHub repository before relying on it for commercial use.
Pros & Cons
✅ Pros
- Long-running agent work can keep going with the laptop closed
- When several agents run in parallel, the stalled one is easy to spot
- You keep your existing agent CLIs, so switching costs are minimal
- A single binary with no dependencies makes installing and removing it trivial
- No account signup, no telemetry, and the source code is public
⚠️ Cons
- It is terminal-first, with no GUI application. Newcomers to the CLI may find it unfamiliar
- Keyboard shortcuts use their own
ctrl+b-prefixed scheme, which needs adjustment if you also run tmux - Released in March 2026, it is a young project where features and behavior still change quickly
- Development is driven by a very small team, so it is a weaker fit where enterprise support is required
- Windows behavior can vary by environment, so test before adopting it there
Comparison with Similar Services
| Criteria | Herdr | tmux | Zellij | Claude Squad |
|---|---|---|---|---|
| Main purpose | Multiplexing coding agents | General terminal multiplexing | General terminal multiplexing | Managing parallel agent runs |
| Agent state display | Sidebar showing working / blocked / done | None | None | Session list |
| Session persistence | Held by a server, reattachable | Yes | Yes | Relies on tmux |
| Extensibility | Socket API and plugins | Config files and plugins | Plugins (WASM) | Configuration-centric |
| Implementation | Single Rust binary | C | Rust | Go |
Who Is It For
- Developers who use Claude Code or Codex daily and run several tasks at once
- Anyone who wants long agent runs to continue while they are out or asleep
- People who prefer to run agents on an always-on machine or rented server rather than a laptop
- Those who avoid Electron apps and account signups and want a tool that stays inside the terminal
- Developers who want to drive agent startup and monitoring from their own scripts via the Socket API
Summary
Herdr addresses two problems that appear once you start running several coding agents at the same time: not knowing which one has stalled, and losing work when the terminal closes. Because it does not replace your existing agent CLIs, the risk of trying it is low ─ it is a single binary you can delete if it does not suit you. It is a young project that still moves quickly, so a sensible approach is to try it on a personal development machine first and roll it out more widely once you are confident it holds up.