agents-libagents-lib
AgentsMCP ServersWorkflowsTutorialsComparisonsTemplatesNewsAbout
HE

herdr

⭐ 11,753•Rust•Agent Infrastructure

Terminal-based agent multiplexer for running multiple AI coding agents simultaneously.

GitHubWebsite →
RustTerminalMultiplexerAgent Orchestrationtmux

Overview

herdr is a terminal-based agent multiplexer written in Rust, designed for running and managing multiple AI coding agents simultaneously. It combines tmux's persistence with agent-specific state awareness — showing whether each agent is blocked, working, done, or idle at a glance. Over 11,000 GitHub stars.

Features

  • ✓Real terminal per agent (not emulation)
  • ✓Agent state visibility (blocked, working, done, idle)
  • ✓Workspaces, tabs, panes for repo organization
  • ✓Persistent sessions across terminal closures
  • ✓Single ~10MB Rust binary
  • ✓Scriptable via local socket API

Installation

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

Pros

  • +Lightweight (~10MB binary, no Electron wrapper)
  • +Agent-specific state tracking
  • +Persistent sessions across reboots
  • +Runs over SSH

Cons

  • −Windows support still in beta
  • −Newer project, less mature than tmux
  • −Learning curve for tmux users
  • −AGPL-3.0 license (commercial use requires separate license)

Alternatives

tmux →Orca →Agent Browser →

Documentation

herdr

Overview

herdr is a terminal-based agent multiplexer written in Rust, designed for running and managing multiple AI coding agents simultaneously. It combines tmux's persistence with agent-specific state awareness — showing whether each agent is blocked, working, done, or idle at a glance.

With over 11,000 GitHub stars and 3,500+ weekly growth, herdr addresses a real operational pain point: as developers run more concurrent AI agents (Claude Code, Codex, Cursor in parallel), they need a way to monitor them without leaving the terminal.

Features

  • Real terminal per agent — each agent gets its own screen, not emulation
  • Agent state visibility — sidebar shows 🔴 blocked, 🟡 working, 🔵 done, 🟢 idle
  • Workspaces, tabs, panes — organize by repo; click, drag, and split
  • Persistent sessions — close the laptop, nothing dies; reattach from another terminal
  • Single ~10MB binary — runs over SSH, no Electron wrapper
  • Scriptable — local socket API and CLI for orchestration
  • Agent-aware — built-in understanding of AI agent states and lifecycles
  • Mouse-native interface — click, drag, and manage with mouse

Installation

Via Install Script

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

Via Homebrew

brew install herdr

Other Options

  • mise (version manager)
  • Nix package manager
  • Windows preview via PowerShell

Quick Start

# Start a new workspace
herdr

# Launch an agent in a new pane
herdr run claude-code --project ./my-app

# Split the current pane
herdr split

# Switch between agents
# Use mouse or keyboard shortcuts

# List all running agents
herdr list

# Reattach to a previous session
herdr attach

Core Concepts

Agent Multiplexing

herdr provides a terminal multiplexer experience specifically designed for AI agents. Each agent gets its own real terminal session with full TTY support.

State Awareness

Unlike tmux, herdr understands agent states:

  • 🔴 Blocked — waiting for user input or external resource
  • 🟡 Working — actively processing a task
  • 🔵 Done — completed the current task
  • 🟢 Idle — ready for new instructions

Workspaces

Organize agents by repository or project. Each workspace can have multiple tabs and panes for different agents.

Persistence

Sessions persist across terminal sessions and even reboots. Close your laptop, SSH back in, and reattach — nothing is lost.

Advanced Features

Socket API

Script agent management via local socket:

herdr api status  # Get status of all agents
herdr api send agent-1 "fix the auth bug"  # Send command to specific agent

SSH Support

Run agents remotely and manage them from any terminal:

ssh user@remote-server
herdr attach  # Reattach to remote session

Integration with AI Tools

herdr works with:

  • Claude Code
  • Codex
  • Cursor
  • Windsurf
  • Any terminal-based AI agent

Examples

Multi-Agent Development Session

# Start workspace for a project
herdr workspace my-project

# Launch Claude Code for feature development
herdr run claude-code --task "implement login feature"

# Split and launch Codex for code review
herdr split
herdr run codex --review

# Split again and launch Strix for security scan
herdr split
herdr run strix --target ./

Parallel Agent Execution

# Fan one prompt across multiple agents
herdr fanout "refactor the auth module" \
  --agents claude-code,codex,cursor

# Compare results and merge the winner
herdr compare
herdr merge agent-1

CI/CD Integration

# Run agents in headless mode for CI
herdr headless --agents "claude-code,codex" \
  --task "run tests and fix failures" \
  --timeout 30m

Pros

  • ✅ Lightweight (~10MB binary, no Electron wrapper)
  • ✅ Terminal-native experience
  • ✅ Agent-specific state tracking
  • ✅ Persistent sessions across reboots
  • ✅ Runs over SSH
  • ✅ Scriptable via socket API
  • ✅ Mouse-native interface
  • ✅ Written in Rust for performance

Cons

  • ❌ Windows support still in beta
  • ❌ Newer project, less mature than tmux
  • ❌ Learning curve for tmux users
  • ❌ Requires terminal access (no web UI)
  • ❌ AGPL-3.0 license (commercial use requires separate license)

When to Use

herdr is ideal when you:

  • Run multiple AI coding agents concurrently
  • Need to monitor agent states without leaving the terminal
  • Want tmux-like persistence with agent-specific features
  • Prefer a single binary over Electron/GUI apps
  • Work over SSH or in remote environments
  • Need scriptable agent orchestration

It positions itself as "tmux rebuilt for agents" — offering terminal multiplexing with built-in agent awareness. Unlike GUI agent managers (which are often Mac-only and closed-source), herdr is terminal-native, cross-platform, and agent-aware.

Resources