An AI that conducts other AIs. The most capable model plans and delegates; cheaper models execute; a fresh judge and a plain script decide what ships. Nothing goes out unattended.
A constitution of hard laws, a deterministic gate that holds the final vote, and sensitive work that queues for a human instead of guessing.
Given a real task, it scouted, planned, built, verified, and gated the change end to end, and shipped nothing unattended.
Clone it, run one command, point it at your own repo. Standard library only. No lock-in, no service to sign up for.
Built on the public OsmoSync method, adopted not reinvented. You get the whole thing to run and adapt.
Sign up and you receive the whole thing, ready to run. Not a waitlist for a product, the actual working system.
The public maestro-starter repo, yours to clone. The code lives on your machine, not behind a login you have to maintain.
Triage, Conduct, Build, Verify, then a deterministic gate, then a human ships. It arrives assembled, not as a diagram you build yourself.
One prompt that puts the conductor to work. Copy it, paste it into Claude Code, and the loop starts on your first task.
A short read that takes you from zero to conducting. No jargon, no setup maze, no video course to sit through.
No service to sign up for, no dependency to install, no vendor to trust. Run it, read every line, change what you want.
Clone the repo and run one bootstrap command. That is the whole install, no extra tooling to chase down.
Drop the kickoff prompt into Claude Code. The conductor wakes up, takes the brief, and calls the loop.
Aim it at your own project and approve what ships. A human signs off on every change before it lands.
MAESTRO is a disciplined autonomous orchestrator, adopted from the public OsmoSync method, not a magic box. It plans, builds, and checks its own work, and a human still approves every ship.
Drop your email and I'll send you MAESTRO the moment it's packaged to install: the repo, the setup, and the exact prompt to start conducting.