anthropics/launch-your-agent

launch-your-agent is a Claude Code skill that walks a technical founder through the process of building an AI agent, from an initial idea to a live, deployed agent running in their own Anthropic account. It handles the full lifecycle: interviewing you about what you want, scoping a v0 version, launching it, grading it against your own definition of done, iterating, and optionally scheduling it to run on a recurring basis. At a high level, it works as a guided conversation inside Claude Code. You start by running the skill with a slash command, and it asks you questions about what you want your agent to do. Based on your answers, it scopes out a minimal first version, creates the agent in your Anthropic account, tests it, and gives you a grade against criteria you defined. If the agent needs to run on a schedule, say, every morning at 9am, it sets that up too. Everything happens through your own API key, so the agent lives in your account and costs you cents per run. This is designed for technical founders who want to build something on Claude's Managed Agents platform but don't want to manually figure out the API, configuration, and deployment details. Concrete examples: an internal worker that summarizes customer feedback, a piece of your product that handles user questions, or a customer-facing agent that drafts email replies. The skill gives you a structured path from "I have an idea" to "there's a working agent in my console" without requiring you to read all the API docs first. When you're done, you walk away with a live agent plus a folder containing your build sheet, the exact API calls used, a reusable launch script, an eval scaffold, and a document outlining what to build next. A companion skill called /wrap-up regenerates an overview page and suggests the next upgrades. The project is a reference implementation, not actively maintained.