AgentManager

About AgentManager

AgentManager is a native macOS application that monitors multiple local Claude Code sessions. It uses a floating window to display the status of each session and surfaces automatically when a session requires user input. The tool relies on the official hook system of Claude Code to track state without scraping terminal output.

Review

Developers running parallel AI coding sessions often miss prompts when switching contexts. AgentManager addresses this through state tracking and alerts the user only when explicit human intervention is required. The application focuses on local workflows and integrates directly with specific terminal environments.

Key Features

• Tracks Claude Code session states using official lifecycle hooks and FSEvents.
• Surfaces a floating macOS window when a session needs input, hiding it when all sessions are clear.
• Jumps to the exact terminal target, including iTerm2 panes, Terminal tabs, and VS Code or Cursor windows.
• Displays session states using pixel art cats in a virtual room, with an option for a simple text mode.
• Plays a cat meow audio alert for waiting sessions instead of using standard Notification Center banners.

Pricing and Value

AgentManager requires a $4.99 monthly or $48 yearly subscription after a 7-day free trial that does not require a credit card. The application costs around 2.5 MB and processes all session data locally on the Mac. Its value lies in preventing idle time for developers who run multiple AI coding sessions simultaneously.

Pros

• Detects session states using official Claude Code hooks, avoiding false positives from terminal output scraping.
• Maintains a stable list order based on launch time, preventing spatial confusion when multiple sessions require attention.
• Keeps all session data and prompts strictly on the local machine without network transmission.
• Includes a global audio debounce that plays a single alert when multiple sessions require input at the exact same time.

Cons

• The tool does not support remote Claude Code sessions running over SSH, as the hook writes state files to the remote filesystem.
• It lacks a fallback mechanism for AI coding agents that do not emit official lifecycle hooks, limiting it strictly to Claude Code.
• Developers orchestrating dozens of headless agents or managing fleet dashboards will find this tool unsuited for their scale and workflow.

This application fits developers who run several local Claude Code sessions in parallel and need a reliable way to monitor their progress. It works best for those using supported terminals like iTerm2, VS Code, or Cursor, and who prefer local-only data processing.

Open 'AgentManager' Website