The rise of AI coding assistants has created a new kind of workflow, but also a new kind of chaos. Running a single agent is one thing; trying to manage several at once—each with its own task, context, and potential to wreak havoc on your codebase—quickly becomes an exercise in frustration. Terminals get cluttered, it's hard to track which agent is waiting for input, and running them in parallel across different git branches is a recipe for disaster. Agent of Empires (AoE) steps into this chaos not just as a tool, but as a command center, designed to bring order and observability to the multi-agent developer.
At its core, AoE is a session manager built on the rock-solid foundation of tmux. This is a brilliant design choice. It means your agent sessions are persistent, surviving SSH disconnects, terminal crashes, or just closing your laptop lid. You can fire up aoe, and all your agents are exactly where you left them. This alone is a significant quality-of-life improvement.
A Tale of Two Dashboards
The primary way to interact with AoE is through its Terminal User Interface (TUI). It provides a clean, at-a-glance view of all your running sessions, their status (running, idle, needs input), and the commands they're executing. From here, you can create new sessions, attach to them to provide input, and review their work. For anyone comfortable in the terminal, it's an intuitive and powerful way to work.
But the killer feature is the web dashboard. By running aoe serve or simply pressing R in the TUI, AoE exposes a web interface that mirrors the TUI's functionality and then some. This untethers you from a specific terminal session. You can monitor your agents from a browser on another machine, your tablet, or even your phone. The remote access is thoughtfully implemented, with options for secure tunneling via Tailscale Funnel or Cloudflare Tunnel, complete with QR code and passphrase authentication.
The web UI offers two distinct views. The first is a raw terminal rendering, which perfectly replicates the tmux session in your browser. The second, more innovative, is the "structured view." This mobile-first interface uses the Agent Client Protocol to render an agent's state natively, with dedicated panels for its plan, cards for tool calls, and even a "swipe-to-approve" mechanic. It's a forward-thinking feature that makes interacting with agents on a small screen not just possible, but genuinely pleasant.
Power-Ups for the Pro Coder
Agent of Empires is packed with features that show a deep understanding of a developer's real-world needs. The integration with git worktrees is a standout. It automates the process of running different agents on separate branches of the same repository, keeping their work isolated until you're ready to merge. For more complex scenarios, it even supports multi-repo workspaces, allowing a single agent session to operate across several local repositories.
Security and isolation are also first-class citizens. The ability to sandbox agents in Docker, Podman, or Apple Containers is critical. When you're letting an AI execute arbitrary code, containing it is non-negotiable, and AoE makes this easy to configure. The extensive list of supported agents—from Claude Code and Gemini to various open-source models—ensures broad compatibility with whatever tools you're already using.
Further customization is available through project-specific configuration files, hooks for automation, and the ability to override agent commands with custom scripts or wrappers. Features like a built-in diff viewer, sound and push notifications, and session resumption for Claude conversations round out a remarkably complete and polished toolset.
The Fine Print
No tool is perfect, and AoE has its limitations. The most significant is its lack of Windows support; it is strictly a Linux and macOS affair. Its reliance on tmux is a strength for those familiar with it, but it does introduce a small learning curve for newcomers who will need to at least learn how to detach from a session (Ctrl+b d).
The project has over 100 open issues on GitHub. While the development team is clearly active—the latest release was just days ago—this number suggests users may encounter rough edges or bugs. Additionally, the explicit mention of "Session Resume" for Claude implies this handy feature might not be available for the many other supported agents.
Final Verdict
Agent of Empires is a masterclass in building a tool that solves a modern, complex problem with elegance and power. It transforms the messy, ad-hoc process of managing multiple AI agents into a streamlined, observable, and remotely accessible workflow. The combination of a robust tmux backend, a functional TUI, and a superb web dashboard makes it indispensable for its target audience. While Windows users are left out and some niche features may have limited agent support, for any developer on Linux or macOS who is serious about leveraging multiple AI coding assistants, Agent of Empires is a must-have. It’s a professional-grade tool that feels built to last.