Launch Your Agent

What it does

• One skill walks a technical founder from interview to a live agent on Claude Managed Agents (CMA), in 4 phases: interview → stage & launch → grade & iterate → run without you
• Grades the agent against your own definition of done and iterates; recurring tasks go on a scheduled deployment, with everything bigger laid out as an explicit v1/v2 plan
• Companion /wrap-up skill refreshes the overview page, recaps every primitive you own, and suggests the next 1–2 upgrades
• Ships reference docs alongside: interview-to-config mapping, an examples bank, and a my-agent/ folder you keep (build sheet, launch script, eval scaffold, NEXT-DIRECTIONS.md)

When to use it

A technical founder shipping an internal worker, a piece of their product, or a customer-facing agent on CMA. Doubles as documentation of what Anthropic considers a well-scoped agent.

Getting started

git clone https://github.com/anthropics/launch-your-agent.git
cd launch-your-agent
claude

Then type /launch-your-agent — skills live in .claude/skills/ and load automatically. An Anthropic API key for your own account is created during the flow (platform.claude.com → API keys) and stored in a local .env file; runs cost cents.

Things to know

Labeled a reference implementation — not maintained and not accepting contributions. Deliberately token-intensive: because it explains each step, it costs more tokens than a purpose-built agent would.