Maestro is a coordinator-driven multi-agent delivery framework. The maestro-* folders below are
ready-to-copy workspaces: copy one anywhere, put your source code inside, run claude (or codex)
in it — the agent follows that template's workflow, gates, and identity.
| Folder | Use it when you want to… | Methodology | Skills |
|---|---|---|---|
maestro-sdlc/ |
Build software end-to-end: BA → design → UI/UX prototype → code → real-user QC | spec-driven-development | 231 |
maestro-adlc/ |
Build AI products/agents with an eval gate before done | eval-driven-ai | 132 |
maestro-enterprise/ |
Operate governed/production agents: compliance, audit, accountability | enterprise-agent-governance | 166 |
maestro-lite/ |
Ship a small tool/prototype fast, minimal ceremony (no specialist advisors) | risk-based-routing | 39 |
maestro-brownfield/ |
Maintain an EXISTING project: deep onboarding, precise tasks, ask-don't-infer | risk-based-routing | 231 |
cp -R maestro/maestro-adlc ~/work/my-ai-app # copy the template anywhere
cd ~/work/my-ai-app # move your service code into services/ or apps/
claude # /coord to start · /ship for autonomous build-to-doneEach template is self-contained and standardized: entry points (CLAUDE.md, AGENTS.md,
COMMAND.md), the .maestro/ workflow brain, a purpose-fit agent/skill set, and folders for your
code and working artifacts. Ask the agent "who are you" — it answers as that Maestro variant
operating your project.
platform/ holds the framework source and the generator. Maintain Maestro there:
cd platform && python3 scripts/build-variant.py --all # rebuild all maestro-* templatesTemplate manifests and the build contract: platform/variants/README.md.
MIT — see LICENSE.