Approve from the notch
Crew Tower blocks on a local socket and sends your choice back into the exact prompt the agent asked with. The answer is resolved in place, so you do not have to type into a terminal window you cannot see.
Crew Tower
Crew Tower is a macOS menu bar app to monitor and approve AI coding agents from the notch, with local prompts, transcripts, and approvals on your Mac.
Product overview
Crew Tower is a Mac menu bar app for watching and approving AI coding agents without leaving the work you are already doing. It keeps each agent’s status, progress, and remaining context visible in the notch, so you can see which session needs attention and act on it without switching into the terminal.
The product is built for local agent workflows. It reads session data from the files and hooks your agent already uses, lets you approve commands or answers from the notch, and returns the response straight back to the agent over a local socket. It supports multiple coding agents at once and aims to reduce the context loss that comes from checking each terminal manually.
Everything it does
Flag risky commands
Dangerous shell commands are identified before you act. Destructive requests are called out on the card so you can make a deliberate allow or deny decision.
Answer agent questions
When an agent asks a question instead of a permission, the options appear as buttons and the selected answer is passed back in the form the agent expects.
Choose how long approval lasts
You can allow once, allow for the rest of the session, or always allow. The always-allow option writes a durable rule into the project’s own settings when supported.
Monitor sessions at a glance
Each session card shows status, elapsed time, context usage, and the agent’s current request. The bar also surfaces subscription usage for Claude or Codex when enabled.
Jump to the right terminal
Clicking a session brings the correct terminal tab forward by matching the title the agent wrote for itself. The app also recognizes tmux panes and several terminal apps.
Where Crew Tower fits
Managing multiple concurrent agents
Use Crew Tower when you are running more than one coding agent and need a single place to see which one is blocked, which one has finished, and which one is still working through its context window.
Responding to agent prompts quickly
If an agent asks for permission to run a command or make an edit, you can approve, deny, or answer the request from the notch without jumping into a terminal tab.
Answering multi-choice agent questions
When an agent asks a structured question with multiple choices, the options appear as buttons and your response is routed back in the format the agent expects.
Finding the right terminal session
If you switch between several projects or terminal tabs, Crew Tower can jump to the exact session that needs attention, even when multiple tabs share a directory name.
Keeping agent activity present but unobtrusive
For screen sharing or distraction-sensitive work, privacy mode keeps the panel shut and sound off while the crew continues running locally.
Pros and Cons
Pros
- Keeps agent approvals and questions visible in the notch instead of forcing terminal switching.
- Handles multiple agents and surfaces which one is waiting, finished, or running out of context.
- Supports several approval modes, including one-time, session-long, and durable project rules where available.
- Works locally on the Mac, with the privacy page stating that code, prompts, and transcripts do not leave the machine.
- Can jump directly to the relevant terminal tab or pane when you click a session.
Cons
- Supported agents and terminals are listed, but the site does not provide a full public compatibility matrix for every possible CLI or terminal.
- The pricing page was not available in the collected sources, so plan details and exact pricing could not be verified from the website snapshot.
- The product is macOS-only, with the site specifying macOS 15 or later.
FAQ
How does Crew Tower work?
Crew Tower is set up on macOS and reads the session data your coding agents already write locally. It monitors agent activity, shows prompts and options in the notch, and sends approvals back over a local Unix socket.
Which agents and terminals does it support?
Yes. The home page says it works with Claude Code, Codex, Cursor, Gemini CLI, Qwen Code and opencode, and it also detects other agent CLIs while they run. It can jump to the matching terminal tab or tmux pane where the session lives.
Does it require a Mac with a notch?
Crew Tower is built for Mac and the site says macOS 15 or later is required. It presents itself in the notch on built-in displays, or as a floating pill on external monitors, and it can be hidden when nothing needs you.
Does Crew Tower send my code or transcripts to the cloud?
The privacy page says the app keeps code, prompts and transcripts on your Mac. Network requests are limited to optional Claude usage checks, license activation and update checks.
Is there pricing information on the site?
The site shows a download for Mac and early-bird pricing, but the pricing page itself is not available from the collected sources. The changelog and privacy pages confirm it is a paid app with license activation handled through Polar.
Quick Facts
- Category
- Developer Tool
- Platform
- macOS 15 or later
- Primary use
- Approve and monitor AI coding agents from the menu bar/notch
- Supported agents
- Claude Code, Codex, Cursor, Gemini CLI, Qwen Code, opencode
- Website
- crewtower.app
- Local data handling
- Code, prompts, and transcripts stay on the Mac
Crew Tower Alternatives
ByteAsk
ByteAsk is a terminal-first AI coding agent for C and C++ that edits repos and verifies changes with compilers, debuggers, sanitizers, and tests.
CreateOS Sandbox
CreateOS Sandbox is an isolated compute environment for running code and agent workloads in Firecracker micro-VMs with private networking and SDK, CLI, or MCP control.
hob
hob is an independent workspace for coding agents, with local control over sessions, terminals, history, routing, and follow-up work.
Manta AI
Manta AI is an autonomous web app testing tool that maps app behavior, catches regressions, and generates tests from a URL, no scripts or selectors needed.
Redline
Redline is a budgeting tool for Claude Code that paces sessions by time, tokens, cost, or plan percentage, helping them end with a usable result.
SonOf
SonOf connects to your repo and PM tool, audits your codebase, and turns approved work into shipped tickets with senior engineering review.