AgentManager

Product Introduction

1. Definition: AgentManager is a native macOS desktop application (built with Swift) designed as a productivity and session management tool for developers and engineers using Claude Code, the command-line interface for Anthropic's Claude AI. It functions as a system-level overlay and status monitor.
2. Core Value Proposition: It exists to solve the critical visibility and context-switching problem inherent in managing multiple, concurrent Claude Code AI coding sessions. Its core value is providing real-time session awareness and one-click terminal navigation, ensuring developers never miss a session requiring input and can instantly resume the correct workflow.

Main Features

1. Global Session Status Dashboard: A persistent, floating window that displays a real-time list of all active Claude Code sessions. Each entry shows the session's status (Waiting, Running, Done, Idle) and duration. The window intelligently surfaces when any session enters a "Waiting" state and hides when all are clear, minimizing screen clutter.
2. Precision Terminal & IDE Navigation: Upon clicking a session in the list, AgentManager performs a context-aware jump to the exact terminal or IDE window where that session is running. It leverages macOS Automation permissions for precise targeting: jumping to the specific pane in iTerm2, the tab in Terminal or Ghostty, or the project window in VS Code, Cursor, or JetBrains IDEs.
3. Keyboard-First Workflow & Menu Bar Integration: The entire interface is designed for efficiency. The main window can be toggled with a customizable global hotkey (default: ⌥ Space). Users can navigate the session list and trigger jumps entirely via keyboard. A complementary menu bar icon provides at-a-glance status via colored dots representing session counts, offering critical information without opening the main window.

Problems Solved

1. Pain Point: "Session Blindness" in Multi-Session Workflows. When running multiple Claude Code sessions across different terminals or projects, it's easy to miss which one has paused and is awaiting user input, leading to significant workflow delays and inefficiency.
2. Target Audience: Software developers, DevOps engineers, and technical leads who utilize Claude Code for iterative coding, debugging, or system operations. Specifically, power users who engage in parallel problem-solving, such as a full-stack developer debugging both backend API and frontend UI issues simultaneously in separate Claude sessions.
3. Use Cases: A developer refactoring a monolith into microservices can have a separate Claude Code session for each service module. AgentManager ensures they see instantly which module's code generation is complete and waiting for review. A data scientist running multiple data analysis scripts with Claude can jump immediately to the session that has generated a result requiring validation.

Unique Advantages

1. Differentiation: Unlike basic terminal multiplexers or manual window management, AgentManager is purpose-built for the Claude Code workflow. It does not just list processes; it understands Claude's specific session states and integrates directly with popular terminal emulators and IDEs for precise, semantic navigation, a feature absent in generic tools.
2. Key Innovation: Its hook-based, non-intrusive detection system. Instead of requiring a custom CLI or complex setup, it registers a lightweight hook directly into the user's Claude Code configuration (~/.claude/settings.json). This allows it to passively monitor all session lifecycles with zero runtime overhead or modification to the claude command itself, ensuring broad compatibility and ease of use.

Frequently Asked Questions (FAQ)

1. Is AgentManager compatible with tmux or terminal multiplexers? Yes, AgentManager can detect Claude Code sessions running inside tmux. However, its precision jump-to-terminal feature currently supports detection only for tmux; it cannot yet navigate to a specific tmux pane or window. Full jump support for multiplexers is a potential future enhancement.
2. What macOS permissions are required for the jump-to-terminal feature to work? The permissions required depend on the target application. For iTerm2, Terminal.app, and Ghostty, macOS will prompt for Automation permission on first use. For Android Studio and similar JetBrains IDEs, Accessibility permission is required. No special permissions are needed for VS Code, Cursor, Zed, Warp, kitty, WezTerm, or Alacritty.
3. How does AgentManager handle my data and privacy? AgentManager operates with a strong privacy-first model. All session data—including prompts, file paths, and session names—is processed and stored locally on your Mac and never transmitted externally. The application only sends anonymized, aggregated crash reports and basic usage statistics for improvement purposes, explicitly excluding any sensitive information.
4. Can I use one AgentManager license on multiple Macs? Yes. The license is per-user, not per-machine. You can install and activate AgentManager on all your personal macOS devices (e.g., a work iMac and a personal MacBook) using the same license key. Each activation is managed independently, and you can deactivate a device from the app's Settings at any time.
5. What happens after the 7-day free trial ends? After the trial, you must purchase a subscription to continue using AgentManager. The trial includes all features. If you do not purchase a plan, the app will cease to function. Subscription options are a Monthly plan at $4.99 or an Annual plan at $48/year (effectively $4.00/month), offering a discount for annual commitment.