agent-manager

Agent Manager is a terminal-based tool for developers who run multiple AI coding agents at once. Instead of juggling separate terminal tabs or windows for each agent session, this tool gives you a single interface where you can start agents, watch their status, answer blocked ones, and review their code changes — all without leaving the list. It works with any CLI-based coding agent you have installed, like Claude Code, opencode, or codex, and runs them in tmux panes under the hood.

The core workflow is fast and keyboard-driven. Press the spacebar on a group row, type a sentence describing the task, and an agent spawns in seconds — already working in the right directory. The same key can also answer a blocked agent's question directly into its pane. Agents rename themselves once they understand the task, so a screen full of them stays readable. You can group sessions by project path, fold subtrees while still seeing a status summary, and reorder everything to match your mental model. The tool detects six agent states (working, waiting, finished, errored, idle, dead) and for Claude Code it uses hook events for accurate detection. A dedicated review mode (ctrl+r) opens a split-screen diff of changed files, with commenting and refresh as the agent keeps editing.

Agent Manager is built for developers who run multiple agents across several projects and need to keep track of them without context-switching. It's especially useful for code reviews, handling several feature branches at once, or debugging a blocked agent mid-sprint. The project is open source (Apache-2.0) and runs on macOS, Linux, and WSL.