herdr: a terminal multiplexer for AI coding agents with real-time state tracking and persistent sessions

What it solves

Herdr is a terminal multiplexer specifically designed for managing multiple AI coding agents. It solves the problem of tracking the state of various agents running in separate terminals, allowing users to see at a glance which agents are working, blocked, or finished, without needing a heavy GUI wrapper or a proprietary app.

How it works

Herdr runs as a lightweight Rust binary that creates a background server to keep agent sessions persistent. It provides a real terminal environment (not an emulator) for each agent, supporting full-screen TUIs. Users can organize agents into workspaces, tabs, and panes using a mouse-native interface or keyboard shortcuts. It automatically detects agent state (idle, working, blocked, done) using process-name matching and terminal-output heuristics.

Who it’s for

Developers who use multiple AI coding agents (such as Claude Code, Devin, or GitHub Copilot CLI) and want a persistent, terminal-based management layer that works across local machines, remote servers via SSH, and even mobile devices.

Highlights

• Agent State Tracking: Automatically labels agents as blocked, working, done, or idle in a sidebar.
• Persistence: Sessions remain alive in the background after detaching, allowing reattachment from any terminal.
• Lightweight: A single ~10MB Rust binary with no dependencies, no Electron, and no telemetry.
• Remote Access: Native support for remote servers, enabling features like image pasting that standard SSH+tmux combinations lack.
• Agent Orchestration: Includes a local socket API and CLI that allows the AI agents themselves to manage their own workspaces and panes.