github.devotd.
NewCoding agents

agent-of-empires/agent-of-empires v1.20.0 — AI Agent Session Manager

Covers v1.20.0, the latest release of agent-of-empires, a Rust session manager that runs multiple AI coding agents in parallel as tmux sessions on Linux and macOS and lets you manage them from a TUI or the web.

Repositoryagent-of-empires/agent-of-empires ↗

Covers v1.20.0, the latest release of agent-of-empires, a Rust session manager that runs multiple AI coding agents in parallel as tmux sessions on Linux and macOS and lets you manage them from a TUI or the web.

GitHub description: Manage multiple Claude Code, OpenCode agents from either TUI or Web for easy access on mobile. Also supports Mistral Vibe, Codex CLI, Gemini CLI, Pi.dev, Copilot CLI, Factory Droid Coding.

stars
★ 3.4K
Forks
385
Language
Rust
License
MIT
Latest release
v1.20.0
Read original source ↗View on GitHub ↗

agent-of-empires/agent-of-empires is a session manager that brings multiple AI coding agents together in one place on Linux and macOS. The latest publicly released version is v1.20.0. It is written in Rust and follows the MIT license. The repository states that it centers on Claude Code and OpenCode while also supporting Mistral Vibe, Codex CLI, Gemini CLI, Pi.dev, Copilot CLI, and Factory Droid Coding. The goal is to attach to sessions through a TUI or a web browser so you can check on them even from a mobile device.

The project is called Agent of Empires (AoE). It runs agents in parallel across multiple branches and divides workspaces using persistent tmux sessions (terminal sessions that survive a disconnect) along with optional Git worktree and container isolation. You can inspect session state and handle notifications and diff reviews via the TUI, the web, the CLI, or an HTTP API. The repository was created on January 9, 2026 and has 3,371 GitHub stars.

However, the provided materials contain no changelog for v1.20.0. What this release actually fixed cannot be confirmed from the currently available information. In the research record, only the momentum score is filled in, at 2.4 out of 10, while the novelty, tech, and quality scores are empty. There are also no figures comparing performance or stability. If your environment requires running agents concurrently across multiple branches, it is worth a look; if you are seeking it for other reasons, waiting until the changelog surfaces may be the better choice.