YoanWai/agent-manager

What it solves

Managing multiple AI coding agents (like Claude Code, OpenCode, or Gemini CLI) simultaneously can be cumbersome, as it requires switching between numerous terminal tabs and manually tracking which agent is active, stuck, or finished. agent-manager provides a centralized interface to run, monitor, and interact with these agents side-by-side without leaving a single view.

How it works

It acts as a thin management layer over existing CLI-based AI agents. It launches each agent in its own persistent tmux session on a private server, ensuring that your existing configurations, logins, and MCP servers are preserved. The manager provides a unified list view with live status detection for supported tools, allowing you to send prompts to agents via a "quick prompt" feature without needing to attach to the session fully.

Who it’s for

Developers who use multiple AI coding agents and want a more efficient workflow for managing concurrent tasks, reviewing changes, and coordinating between different AI tools.

Highlights

• Multi-Agent Orchestration: Run multiple agents (e.g., Claude Code, Grok, Gemini CLI) side-by-side in a project tree.
• Integrated Diff Review: Review changes made by agents as full-file diffs with the ability to leave comments that are sent back to the agent as review prompts.
• Quick Prompting: Send messages to any agent using a single keypress (space) without fully attaching to the session.
• Session Management: Ability to kill, revive, restart, or fork conversations into separate named sessions.
• Git Worktree Integration: Option to spawn sessions into their own git worktrees to keep experiments isolated.
• Live Status Detection: Built-in support for tracking the status of popular AI CLI tools to see at a glance which agent is waiting for input or is working.