Herdr
One terminal for all coding agents – like tmux, but agent-aware
AI Summary
Herdr is a terminal-based agent multiplexer that brings together multiple coding agents (e.g. Claude Code, Codex, OpenCode) in a single terminal interface – similar to tmux, but with agent state detection. It runs locally, on servers, or remotely via SSH, without Electron, a browser dashboard, or a cloud account. It targets developers who want to orchestrate multiple AI coding agents simultaneously and monitor them from anywhere.
✓ Pros
- +Runs anywhere SSH works – no GUI, no Electron, no cloud account required
- +Displays agent status (blocked, working, done, idle) at a glance across all sessions
- +Extensible control via CLI and JSON socket API enables full automation
✗ Cons
- −Windows support is still in beta and only available as a preview
- −Steep learning curve for developers unfamiliar with tmux-style terminal multiplexers
Use Cases
- →Monitor and control multiple coding agents in parallel within a single terminal
- →Remote development: run agents on a server and observe them via SSH from your phone
- →Automated agent orchestration via CLI and JSON socket API
- →Persistent sessions: close the terminal and let agents keep running, then reattach later
Who is it for?
Ideal for software developers who work with multiple AI coding agents on a daily basis and want to manage them efficiently from anywhere – including remotely via SSH.
Tags
What is Herdr?
Herdr is a terminal-based multiplexer for coding agents. The core principle will be familiar from tmux: multiple sessions in a single terminal interface, running persistently in the background, accessible from anywhere via SSH. Herdr extends that concept with agent-state detection. The tool knows whether an agent is currently working, blocked, waiting for input, or finished. It runs locally, on a remote server, or in any environment where SSH is available. No Electron, no browser dashboard, and no cloud account required.
Core features
- Agent multiplexing: Multiple coding agents (Claude Code, Codex, OpenCode, and others) run in parallel within a single terminal interface.
- Agent state display: Herdr shows the current status of each agent at a glance: blocked, working, done, or idle.
- Persistent sessions: The terminal can be closed and the agents keep running. Reattach whenever needed.
- Remote access via SSH: Start agents on a server and monitor them from a laptop or phone, with no additional software on the client.
- CLI and JSON socket API: Control and automation are fully scriptable.
Who is Herdr for?
Herdr is aimed at developers who run several AI coding agents in parallel on a daily basis. Anyone who only occasionally starts a single agent does not need it. The value becomes clear once you are running agents on remote servers and want to monitor their progress from wherever you are. Users new to tmux should expect a learning curve: the concepts of sessions, panes, and attach/detach are not self-explanatory. Windows users should be aware that support on that platform is still considered a preview.
Context & alternatives
Herdr occupies a niche that classic terminal multiplexers like tmux or screen do not cover: agent-specific state status is absent from those tools entirely. Browser-based agent dashboards and proprietary cloud interfaces from individual AI vendors point in a similar direction, but require a GUI or a cloud connection. Herdr stays entirely in the terminal and works anywhere an SSH connection can be established. For anyone who wants to orchestrate agents through scripts and accepts no dependency on external services, Herdr offers an approach that generic terminal multiplexers do not.
