For Claude Code and Codex · on your machines
Your machines. Work lands.
Parallel lanes on your subscriptions keep work moving while you're away. An independent review on every change.
From 6¢ per million input tokens. Output is always free.
6¢ on Pro 500. Plans from $20/month.
- Plus$0.25$20 a month
- Pro 200$0.12$200 a month20x the usage
- Pro 500$0.06$500 a month100x the usage
Included usage compared with Plus. Prepaid dollar credit, never a feature bundle.
Separate from your ChatGPT or Claude subscription.
Spending more than $500 a month? Lower rates.
Talk to us- Draft
- Work
- Review
- Land
- Keep going
Your conversation stays separate. A steer sends this lane a new instruction.
Master Orchestrator knows you
Before you start, ask the hard questions.
Choose a question. Try the controls.
02 · Am I allowed to use my subscription this way?
Yes, with an eligible plan. Use your own login.
Yes. Connect your own Claude Code or Codex login and follow your provider’s plan terms and usage limits. API keys use separate provider billing. Choose your subscriptions below.
Sign in through the provider’s own CLI. Keep the account in your name. Check the terms for your plan and the way you intend to run your fleet.
Provider eligibility and usage limits apply to each account.
09 · What about local models?
Run Laya on your own machine.
Laya runs locally. Master Orchestrator charges nothing for local-model tokens. Choose a local model when you want the model’s processing to stay on your machine.
Choose Laya in the source wall below. Hosted providers send model requests to their servers; local models process requests on your machine.
10 · How do plugins work, and what can they touch?
Give your fleet a new skill.
Choose a plugin from the Master Orchestrator store to add instructions for a specific task. Check its author, source, license and service connections before installing. A plugin’s access depends on the tools and accounts you give it.
- Start with the task
- Choose a skill for work you want your fleet to do.
- Know whose work you use
- Check the author, source and license.
- Check the access
- Read what the plugin can run and which services it connects to before adding it.
11 · What does the machine scan learn?
Teach your fleet how you work.
With your permission, Master Orchestrator scans for languages, build tools, commit patterns and signed-in agents. A few questions establish preferences such as how to handle flaky tests. Your answers guide the workers.
- languages and frameworks
- test and build tools
- how you commit
- which agents are signed in
One question. A test is flaky. What should your fleet do?
Illustrative. The real scan runs on your machine, only with your consent.
01 · Can I cap my bill?
Yes. Set a monthly cap.
Open Settings in the console and enter the amount. Master Orchestrator stops starting new work when you reach it. Your providers bill subscriptions and API usage separately.
Separate provider-usage example: a weekly quota ceiling checks the account feed before starting work.
Illustrative provider-usage controls. Set your monthly bill cap in the console.
03 · Where does my code go?
Your code runs on your machines.
Workers run on your machines. Hosted models receive the prompts and code needed for the task through your provider account. Local models process them on your machine. Master Orchestrator billing receives aggregate token totals; prompts and code stay out of billing.
Your machine
Your repo, the workers, and Master Orchestrator.
Anthropic or OpenAI
The model you signed in with. Your plan, your account.
Master Orchestrator billing
Aggregate token totals under the billing policy.
Illustrative trace. Billing rule from the owner's metering decision: exact totals only, no content, no per-request detail.
08 · Can I mix Claude and Codex?
Yes. Pick the model for each worker.
Yes. Connect Claude Code with your Claude subscription and Codex with your ChatGPT subscription. Assign either to a worker. Each change gets an independent review before landing. Try switching the writers below.
07 · Can I set up without Terminal?
Yes. Double-click the installer.
Yes. Open the Mac installer from your invitation. If your Mac needs Python, the installer tells you where to get it. Then connect Claude or ChatGPT from the console in your browser.
Documented path: macOS or Linux. The local console preview also needs Node.js 18+. This browser replay checks no login and runs no AI session.
04 · What if a worker breaks my repo?
Review catches a bug. The worker repairs it.
Each change gets an independent review before landing. When review finds a bug, the worker repairs the change and submits it again. Review can miss defects, so keep human review on production and security decisions. Try the example below.
Change ready · awaiting landing
Illustrative Postgres review with Claude and Codex.
06 · Am I locked in?
No. Keep your repo and every commit.
No. Work lands as ordinary Git commits in your repository. Your provider accounts stay in your name. Uninstalling removes the recorded program files, skills and hooks while preserving your board, reports and existing CLI settings.
Ordinary commits from the simulated fleet above
Uninstall removes
- Master Orchestrator's program files
- The skills the installer added
- The read-only session hook
Stays yours
- Your repos and every commit on main
- Your board and reports
- Your existing Claude Code and Codex settings
- Your plans, billed by Anthropic and OpenAI
Uninstall removes exactly what the installer recorded, and leaves your board and reports in place.
05 · What happens when I press Stop?
Stop tonight, continue tomorrow.
Master Orchestrator saves your place when you stop. Press Continue to resume the stopped work. Landed commits stay in your repository. Try Stop, reload this page, then Continue in the demo above.
Browser demo: , reload this page, press Continue.
12 · Has Master Orchestrator done real work?
Yes. Master Orchestrator builds itself.
Yes. Master Orchestrator runs its own development fleet: workers write changes, independent reviewers check them, and reviewed changes land. Run the browser example to see that sequence.
Explore the worker, review and landing sequence in this browser example.
Browser simulation of the worker, review and landing sequence.
Bring your subscription. Put your fleet to work.
Bring your Claude and ChatGPT subscriptions. Tap the sources you use to build your fleet.
Use your own accounts. Provider terms and usage limits apply.
Add curated skills from the plugin store.
Your worker paths
Selection is illustrative. Your accounts remain with their providers. Master Orchestrator billing is separate.
Got an extra laptop or desktop? Put it to work.
Give each machine a place in your fleet. One can work on tests while another writes a migration. Talk to Master Orchestrator once to steer the whole fleet.
One conversation 1 machine
Illustrative: tap to add machines, × to remove one. Windows machines run workers in WSL. Actual machines need setup and account access.
Master Orchestrator is opening by invitation.
We set up each new fleet with its owner, a few people at a time. Leave your email and we will send your invitation with the signed Mac installer.