Skip to main content

Quick Overview

In this build-along tutorial, Alex Finn demonstrates Herdr, an open-source terminal manager designed for running AI coding agents. Finn explains how the tool replaces standalone desktop client wrappers by coordinating agents across projects and connected machines. The video walks through installation, interface structure, concurrent agent workflows, and custom peripheral interfaces.

Key Points

  • 1.Herdr is an open-source terminal management tool that coordinates multiple AI coding agents across different projects and machines.
  • 2.The application organizes workflows into devices, spaces representing specific repositories or projects, and tabbed terminal sessions for dedicated agents.
  • 3.A background task queue automatically shifts completed agents to the top of an interface list, indicating when they are ready for new prompts.
  • 4.Custom extensions like Herdr HUD provide a persistent floating overlay that enables users to monitor and interact with agents while running other applications or games.
  • 5.The tool supports local and remote machines as well as heterogeneous AI agents, including Codex and Claude Code, simultaneously within a single workspace.

Summary

Alex Finn introduces Herdr, a free and open-source terminal application designed to streamline workflows with autonomous AI coding agents. Finn explains that he has abandoned standalone desktop wrappers for Claude Code and Codex in favor of Herdr. The tool serves as an interactive, clickable terminal manager that orchestrates multiple autonomous agents running concurrently across different local and remote devices, allowing users to sustain high development throughput without remaining actively tied to a single command-line session.

The Herdr interface organizes work into a hierarchy of machines, spaces, and agent tabs. Finn shows his setup connected to a Mac Studio and a remote laptop, with separate spaces designated for individual software repositories, including newsletters and game prototypes. Inside any selected space, users can open individual tabs representing distinct agents. Finn demonstrates this by assigning one agent to core mechanics, another to user interface design, and others to audio or progression systems. These tabs can independently host different agent backends, such as Codex, Claude Code, or Gemini. On the left side of the screen, Herdr maintains a unified, chronological list of active agents across all devices. When an agent finishes running commands or code modifications, it jumps to the top of the list with a status indicator, alerting the developer to review output and issue the next prompt.

Installation requires running a single shell script from the project repository, after which launching the herdr command opens the terminal interface. To demonstrate a build from scratch, Finn creates a new directory space and initializes a top-down Three.js roguelike game inspired by Diablo. He prompts an initial Codex agent to produce the baseline combat arena, player movement, enemy waves, and stat upgrades. Once the basic build runs successfully in the browser, Finn branches out work to parallel agents. He opens separate tabs for loot drops, additional enemy classes, larger map geometry with hidden chests, and character class abilities. Because Herdr allows agents to run simultaneously, each sub-system is developed concurrently without blocking the main terminal session.

Finn then details Herdr HUD, an open-source companion overlay he developed to keep agents accessible during other desktop activities. The HUD appears as a minimal floating badge that stays on top of all windows, including fullscreen video games. As background agents finish build steps, the badge increments a notification count. Clicking the badge slides open a modal that displays recent agent logs, interactive terminal views, and prompt input boxes, enabling prompt dispatch without leaving the active game or application. Finn concludes by demonstrating mobile and Apple Watch interfaces tied into the same Herdr backend, allowing users to monitor compilation progress and reply to agent prompts away from their desks.

Terminal Management and Agent Organization

Herdr operates as an enhanced terminal environment designed to coordinate autonomous AI agents. The left sidebar displays connected hardware, such as local desktops and laptops, nested with project spaces. Within each space, users open dedicated tabs for different agents assigned to distinct roles, such as gameplay programming, map generation, 3D modeling, and testing. A chronological activity panel tracks agent statuses, surfacing finished tasks to the top of the queue for rapid prompt iteration.

Parallel Game Development Demonstration

To demonstrate the multi-agent workflow, Finn sets up a project space to build a browser-based 3D roguelike demon-slaying game using Three.js and Codex. After generating the foundational gameplay loop, additional tabs are instantiated for loot mechanics, enemy types, expanded map layouts, and class abilities. Multiple agents run simultaneously across these tabs, allowing features to be written, integrated, and verified in parallel while the core game remains playable locally.

Custom HUD and Mobile Extensions

Because Herdr is open source, users can build custom interfaces on top of its agent infrastructure. Finn showcases Herdr HUD, an overlay plugin that places a draggable icon on top of fullscreen applications and video games. Clicking the icon reveals an interactive drawer showing agent activity and text inputs, accompanied by unread notification counters. Finn also demonstrates custom mobile and wearable interfaces that relay agent statuses and prompt controls to an iPhone and Apple Watch.

The Bottom Line

The video demonstrates how Herdr acts as an orchestration layer for AI coding agents by unifying terminal management, multi-device networking, and parallel workspaces. It establishes that offloading sub-tasks to concurrent agents and tracking them via a unified queue significantly speeds up rapid prototyping. The demonstration leaves unresolved how Herdr handles conflicting code merges across agents touching the same codebase simultaneously, as well as the long-term API costs associated with running multiple persistent models in parallel.

FAQ

What is Herdr and what does the application allow users to do?

Herdr is an open-source terminal manager that allows developers to run, organize, and monitor multiple autonomous AI coding agents across different projects and machines within a single interface.

How does Herdr organize different programming projects across multiple machines?

Herdr groups environments by physical hardware devices, subdivides each device into spaces corresponding to project directories, and runs distinct AI agents within tabbed terminal sessions inside each space.

What is Herdr HUD and how does it function during desktop tasks?

Herdr HUD is a draggable floating overlay plugin that stays on top of other desktop windows and games, alerting the user to agent updates and allowing prompt submission without switching applications.

How does Herdr notify users when an autonomous coding agent completes a task?

The application maintains an agent list where agents that complete their commands automatically jump to the top with an updated status, alongside badge notification counters in the companion HUD overlay.

Which AI models and coding agents are demonstrated running inside Herdr?

The video demonstrates running OpenAI Codex and Claude Code inside Herdr tabs, while noting compatibility with other command-line agents and models such as Gemini.

Worth watching for

Software developers, technical hobbyists, and AI vibe coders seeking an organized, multi-agent terminal workflow to manage autonomous programming tools across several projects and machines.

  • herdr
  • ai-agents
  • vibe-coding
  • codex
  • claude-code
  • terminal