Private Beta · Mac

One chat box for Claude, Codex, Gemini, and Grok.

Route each task to the right model and write straight to your real files. Your subscriptions. Your machine. Your code never leaves it.

Runs 100% on your computer. No API keys, no cloud.
MMaestro
Maestro
ChatReview

Start a local run

Select a working directory, choose Auto or a specific engine, then send a prompt. Runs stay tied to this Mac and the selected folder.

codexo-next
AutoClaudeCodex
Options
Send a prompt... (Enter to send, Shift+Enter for newline)Send

The problem

Two terminals. Two tools. Zero shared memory. All the copy-pasting is on you.

Claude Code in one terminal. Codex in another. They do not know what the other just built. No shared context. No shared memory. Every handoff is a copy-paste, and neither engine can see your project the way a real collaborator would. Maestro puts both under one roof, tied to your files, with context that carries.

Claude Codeterminal 1
Codexterminal 2
no shared memorycontext lost
Neither engine knows what the other just built

How it works

Connect once. Work like it's one tool.

01

Install.

When the Mac app ships, install in one click. No terminal required.

02

Connect your CLIs.

Point Maestro at the Claude Code and Codex CLIs already installed on your machine. It uses the auth sessions they already have. Nothing new to sign into.

03

Just work.

Type in one box. Maestro routes each task to the model that handles it best, edits your local files, and remembers the project. Switch models by hand any time, or let it conduct.

Active run

This is what conducting looks like.

A project list on the left, the active run in the center, an engine picker, and a composer that routes each task to Claude Code or Codex. It all runs on your machine, against your real files.

Maestro · local runtimeopen

The desktop app, PWA, and local server are the same runtime.

Privacy

Your keys. Your machine. Your code.

Maestro runs entirely on your computer. Claude Code and Codex authenticate locally using the CLIs you already have installed. Your files stay local, and your prompts go straight from your machine to the engines you already run. We never proxy your traffic, store your credentials, or see a single line of your work. If our servers went down tomorrow, your local cockpit would keep running.

No code sent to us, ever.

No API keys handed over. You stay signed in to your own accounts.

Works offline-first. The app is yours, not a rented window into ours.

Data flow

Your Mac

Maestro app

Your project files

Your subscriptions

Direct connection (encrypted)

AI Engines

Claude Code

Codex

Codexo servers
(zero access to your data)

Engines

Claude and Codex today. More engines when they are real.

The cockpit is wired for Claude Code and Codex, with Auto routing and a manual override in the composer. Other engines land here when their local workflow is ready, not before.

Claude Code

Anthropic

Live

Codex

OpenAI

Live

Gemini

Google

Soon

Grok

xAI

Soon

Gemini, Grok, and more join as their local tooling matures.

What it does

Built for how you actually work.

01 Route

Knows which engine to run.

Reasoning-heavy work goes to Claude Code. Fast, mechanical edits go to Codex. Auto picks for you, or override with one tap.

02 Edit

Writes to your real files.

Not a sandbox, not a copy. Maestro works in your actual project directories, the same folders already open in your editor.

03 Remember

Never starts blind.

Switch from Claude Code to Codex mid-task and the next engine picks up with full project context instead of starting over.

Run several at once.

Fire off parallel tasks in different projects and watch them work side by side.

Reach it from your phone.

Your cockpit, private to you, on any device. Kick off a run from the couch; it runs on your Mac.

What's coming

Today it conducts your models. Next, it conducts everything.

We ship the cockpit first and build the rest with the people in the Discord.

NowLive

The local cockpit

Claude Code and Codex in one workspace. Auto routing, real file edits, and project-scoped memory. Running today.

Next

More engines

Gemini, Grok, and others join as their local tooling is ready to drive from your own machine.

Later

Connected tools and memory

Bring your own tools and data into one memory layer the agents can draw on across every run.

Later

Named, reusable agents

Configure an agent once for recurring work, then run it on demand from the cockpit.

Founder access

Get in before launch. Lock your price for good.

Maestro is in private build for Mac. Join the waitlist free, and you're first in line when the Mac app ships. Want your seat guaranteed? Put down a $20 refundable deposit — it credits toward your plan — and lock Maestro Pro at $15/mo for life: half off the $30 price, forever. Change your mind before launch and we refund it, no questions.

Waitlist

Free · Private beta for Mac

Be first to get the Mac app

Founder Discord access

By joining, you agree to receive automated and AI-generated updates from Maestro and accept our Terms and Privacy Policy. Consent is not a condition of purchase.

Limited

Founding Member

$20 refundable · credits to your plan

Everything in the waitlist, plus:

Maestro Pro locked at $15/mo for life — 50% off, forever

Your $20 credits toward your plan

Priority onboarding when you get access

Fully refundable until launch

You already pay your AI providers directly. Maestro is the tool that makes them work as one. We never charge for tokens or take a cut of your usage. At launch: Free tier (one engine) and Pro at $30/mo.

Community

Built in public, with the people who use it.

Maestro is shaped in our Discord: per-engine setup help, routing recipes people share, and features shipped from real requests. Founding members get a locked-in price and a direct line to the build.

Join the Discord

Maestro Community

Setup help, routing recipes, and a direct line to the build.

FAQ

Common questions.