Agent-Manager is a terminal user interface (TUI) built as a thin abstraction layer over tmux to orchestrate multiple AI coding agents simultaneously. It enables running various CLI tools—including Claude Code, Codex, OpenCode, Grok, Gemini CLI, Pi, Command Code, and Hermes Agent—within individual, persistent tmux sessions under a private "agentmgr" server. This architecture ensures that each agent retains its own authentication, configurations, and MCP servers while remaining isolated from the user's primary tmux environment. The manager provides a unified project tree view with live status detection and advanced interaction capabilities. Users can interact with agents via a "quick prompt" using the spacebar or perform a sophisticated diff review by pressing Ctrl+R, which allows for line-by-line commenting that is sent back to the agent as a review prompt via the C key. Additional features include session forking (F), automated revival of dead sessions (v), and the ability to spawn agents into dedicated git worktrees. Developed for macOS, Linux, and WSL2, the tool supports mouse navigation for selection and scrolling, and provides integrated shell access (T) for builds and Git commands. While it currently lacks cost tracking, it simplifies complex multi-agent workflows by managing session lifecycles and providing a centralized dashboard for monitoring agent progress.