Agentic Command

About This Project

Agentic Command is a desktop command center for interactive AI agent CLIs. It wraps each agent in a real pseudo-terminal, preserving the prompts, permission requests, follow-up questions, and other terminal behavior that make agent sessions useful.

The app is built for the moment when one agent session becomes several. A persistent sidebar launches sessions in chosen working directories, while the main workspace presents every active session as a card. Each card surfaces its current status so it is easy to see which agents are working, waiting for input, or need attention.

What It Does

• Runs multiple concurrent agent sessions, each backed by its own PTY process.
• Keeps session output and exit status available from a central command center.
• Opens a full raw terminal for any session without disrupting the others.
• Highlights permission prompts, questions, idle sessions, and errors.
• Works with Claude CLI by default while keeping the command editable for other terminal-based agents.

How It Works

Agentic Command is an Electron app built around a multi-session manager. The main process owns the PTY processes and routes events by session ID. A preload bridge exposes a focused IPC API, and the renderer maintains the command-center cards, terminal views, output buffers, and attention states.

The result is a practical workspace for supervising several agent-driven tasks without flattening them into non-interactive background jobs.