The Command Center for Coding Agents

One app for Claude Code and your entire git workflow.

Commander interface showing AI-powered code editing, git workflow, and native macOS design Commander interface showing AI-powered code review and design insights

Prompt the agent. See the diff. Commit what's good. That's the whole workflow.

Download for Mac Learn More

Requires a supported agent CLI installation

1. Pick a repo

Open your project. Pick a branch or worktree. That’s your starting point.

2. Prompt the agent

Talk to Claude Code, Codex, or any supported agent. No new accounts, no setup dance.

3. Review & commit

See exactly what changed. Keep what’s good. Commit and move on.

Watch the Loop

Prompt, diff, commit. Here's what that actually looks like.

Agent Profiles

Work account, personal account, local model, cloud model — set them up once, assign to projects, switch instantly.

What It Actually Does

Agents generate code. Commander makes sure you actually review it before it lands.

One App, Any Agent

Claude Code today, Codex tomorrow. Switch agents without touching a config file.

See the Diff Before It Ships

Every agent run is a reviewable unit. No more "wait, what did it change?"

Worktrees That Just Work

Run three tasks in parallel on separate branches. No manual juggling, no stepping on your own work.

Scriptable Actions

Hook up your own scripts to the toolbar. Run builds, linters, deploys — whatever your project needs.

Tabs, Windows, Your Call

Multiple chats in tabs. Detach any conversation into its own window. Arrange things however your brain works.

Quit Without Fear

Close the app, reopen it, pick up right where you left off. Sessions survive relaunches.

SSH Into Anything

Run agents on a remote Linux box over SSH. Review everything locally on your Mac.

Accessible by Default

Full VoiceOver support, clear hierarchy, keyboard-navigable. Not an afterthought.

Not an IDE. Not a terminal replacement. Not another Electron app.

30 MB. Native Swift. A magnitude less than any Electron app. Commander is the review loop your agents are missing.

Not a Toy

This is for the work where "the agent wrote it" isn't an acceptable commit message.

Feature Delivery

Break the work into isolated runs. Review each chunk. Merge when it's actually ready.

Complex Refactors

The kind where one wrong rename breaks 40 files. See every change before it lands.

Migrations

Try two approaches in separate worktrees. Compare. Ship the one that doesn't scare you.

Bug Triage

Prompt, get a patch, read the diff. Commit only when you understand why it works.

Supported Agents

Not locked to one provider. Bring whichever agent CLI you already use.

Claude Code
Codex
OpenCode
Pi

View setup details in documentation

The Problem with Everything Else

You already have tools. Here's why they don't quite work for agent-driven development.

Terminal

  • You're scrolling 400 lines of agent output trying to spot the one file it shouldn't have touched
  • Prompting in one pane, diffing in another, committing in a third
  • Worktrees are powerful — if you enjoy typing git worktree add all day

VS Code

  • Great editor. But your AI workflow is three extensions, two sidebars, and a prayer
  • Hard to treat one agent run as a single reviewable unit
  • Your setup works on your machine. Good luck on your teammate's

Commander

  • One loop: prompt → diff → commit. That's the whole app
  • Native macOS app. Fast, accessible, no Electron
  • Built-in worktree management for parallel tasks
  • Any supported agent, same workflow
  • Tabs, detachable windows, sessions that survive relaunches

Don't Take Our Word for It

Real developers, real workflows, real opinions.

Shannon Potter avatar
Shannon Potter @cifilter
I reached for Commander as soon as I heard about it. Having a native, focused workflow to iterate with AI just works for me in a way that busy editors with a dozen frames don’t.
Vinicius Carvalho avatar
Vinicius Carvalho @viniciusc70
I started using Commander since the first version and saw all the progress and how it evolved. Today it is a tool that is indispensable in my setup when I am creating some software or a feature.
Dominik Drąg avatar
Dominik Drąg @dominikdrag_
I started using Commander out of curiosity. I love the fact that it gives me insight into the agent's thought process and gives me more control. The fact that the app is native is a killer, it's smooth as butter. Finally an AI app that doesn't turn my Mac into a heater.
An Tran avatar
An Tran @antranapp
I have been trying Commander recently since I want a native macOS experience with snappy performance and a clean, native UI. I also really like that it supports multiple AI providers, I can easily switch models without switching between multiple apps.

How It Works

Three steps. No magic. No setup wizard.

Tell It What You Need

Describe the feature, the bug, the refactor — in plain English. Commander sends it to your agent with full project context.

Read the Diff

The agent writes code. You see exactly what changed, with syntax highlighting and inline diffs. Don't like something? Tell the agent to fix it.

Commit When You're Ready

Branch, commit, and push — all from inside Commander. No context-switching to a terminal or git GUI.

Privacy & Data

Commander is a local Mac app. Your keys stay in your CLI. Your code stays on your machine.

Credentials

Commander never sees your API keys or passwords. Your agent CLI handles all authentication — Commander just talks to it.

Your code

Your repos never leave your machine via Commander. When you run an agent, the CLI may send context to its provider — that’s between you and the provider.

Website analytics

This site uses TelemetryDeck for anonymized analytics. No cookies, no PII, no tracking you across the web.

Frequently Asked Questions

What is Commander?

A Mac app that wraps your coding agents in a proper review loop. Prompt the agent, see what it changed, commit what's good. That's the core of it.

Is Commander free?

Yes, Commander is free to use. Just download it and go — no account required.

What macOS version do I need?

macOS 15.0 or later. It's a native Swift app, so it stays current with the latest macOS features.

How does it talk to agents?

Commander connects to the agent CLI you already have installed (Claude Code, Codex, etc.) and adds a review-first workflow on top. No extra accounts, no proxy layer.

What do I need to get started?

A Mac running macOS 15.0+ and one supported agent CLI. That's it. Setup takes a few minutes.

More questions? Check the documentation →

No Extra Moving Parts

You already have the agent CLI. Commander just makes it usable.

No new accounts

If your agent CLI works in the terminal, it works in Commander.

Your repos, your machine

Branches, worktrees, and project state — all visible, all local.

Accessible by default

VoiceOver, keyboard navigation, clear semantics. Built in from day one.

Who Built This

Commander is built by Marcin Krzyzanowski, a Swift developer who got tired of juggling terminals, git GUIs, and agent windows every time he wanted to ship something. He also makes Notepad.exe and DoctorMarvin.ai.

Just Download It

Free to use. Native Mac app. No account required.

Download for Mac
$ brew install commander click to copy

Requires macOS 15.0+ and a supported agent CLI