Crew Tower

Crew Tower is a macOS desktop application that turns the MacBook notch into a live control tower for monitoring and controlling AI coding agents. It displays the status, progress, and remaining context of each running agent directly in the notch, allowing developers to keep an eye on multiple coding sessions without switching away from their current work. When an agent requests approval for a shell command or asks a question, Crew Tower surfaces the request in the notch with the exact command, a risk assessment for destructive operations, and one-click actions to approve, deny, or answer. The decision is sent back to the agent through its original communication channel, eliminating the need to manually type responses in a terminal window. The app also provides quick navigation to the exact terminal tab where an agent is running, shows quota usage for tools like Claude and Codex, and includes playful pixel companions that react to the user's crew. It requires macOS 15 or later and works with any AI coding agent that supports local socket-based hooks, though it is primarily tested with Claude and Codex.

Category: AI Agent
Pricing: Unknown

Problem solved

Crew Tower solves the problem of managing multiple AI coding agents simultaneously, where agents often block silently waiting for approval or finish without notification. It centralizes agent status, requests, and context usage into a non-intrusive notch interface, allowing developers to approve commands, answer questions, and jump to the right terminal tab instantly without losing focus on their primary task.

Use cases

Approve or deny shell commands from an AI agent without switching to the terminal
Monitor the status and progress of multiple coding agents at a glance
Answer multiple-choice questions from agents with one click
Jump directly to the terminal tab where a specific agent is running
Track context usage and quota for Claude and Codex
Review and approve multi-step plans from agents before execution

Audience

Developers
ML Engineers
Data Scientists
Hobbyists

Pros

Non-intrusive monitoring from the Mac notch
Fast approval workflow without leaving current app
Highlights destructive commands for safety
Multiple approval modes (once, session, always)
Shows context and quota usage

Cons

Only available on macOS 15 or later
Pricing is not publicly transparent (early bird pricing)
May not support all AI coding agents
Relatively early-stage software

Official website: https://crewtower.app/

Canonical LambdaBase page: https://www.lambdabase.com/tools/crew-tower

← Directory

Crew Tower

Unknown

• Updated Aug 21, 2026, 09:16 AM

Crew Tower is a macOS desktop application that turns the MacBook notch into a live control tower for monitoring and controlling AI coding agents.

Use Cases ▼

Approve or deny shell commands from an AI agent without switching to the terminal

Monitor the status and progress of multiple coding agents at a glance

Answer multiple-choice questions from agents with one click

Jump directly to the terminal tab where a specific agent is running

Track context usage and quota for Claude and Codex

Review and approve multi-step plans from agents before execution

What Problem It Solves ▼

Key Features ▼

✓

Non-intrusive monitoring from the Mac notch

✓

Fast approval workflow without leaving current app

✓

Highlights destructive commands for safety

✓

Multiple approval modes (once, session, always)

✓

Shows context and quota usage

Overview

Full Description ▼

When an agent requests approval for a shell command or asks a question, Crew Tower surfaces the request in the notch with the exact command, a risk assessment for destructive operations, and one-click actions to approve, deny, or answer. The decision is sent back to the agent through its original communication channel, eliminating the need to manually type responses in a terminal window.

The app also provides quick navigation to the exact terminal tab where an agent is running, shows quota usage for tools like Claude and Codex, and includes playful pixel companions that react to the user's crew. It requires macOS 15 or later and works with any AI coding agent that supports local socket-based hooks, though it is primarily tested with Claude and Codex.