The onboarding prompt

You've connected deploymill to your agent. Now what? Instead of clicking through a setup wizard, hand your agent one prompt and let it drive: it reads the docs, builds a small live example app, shows you the core loop, and leaves you with something running.

This is the fastest way to see what deploymill does. Copy the prompt below, paste it to your connected agent, and follow along.

Copy this into your agent

You have deploymill connected as an MCP server, an agent-safe control plane for shipping web apps. Walk me through it by building a small live example app end to end, narrating each step so I learn how it works.

1. Get your bearings. Fetch the deploymill://docs/getting-started resource and the deploymill://guides index so you know the tools and the happy path, then call get_account to confirm I'm signed in and check my plan, quota, and spend headroom before creating anything.

1. Ship a live app. Call start_project to create a small example web app named hello-deploymill on the node stack, picking a simple starter template. Wait for the build, then give me the live URL and confirm it's healthy with get_app (show me the health block).

1. Show me the core loop on that app, briefly:

1. Leave me with a living app. Do not stop or delete it. Wrap up with: the live URL, a short recap of what you did, roughly what it costs (idle apps sleep and cost nothing), and three things I might try next.

Use the tools yourself. Don't send me to a dashboard to click things. If a step fails, tell me the error code and how you're recovering. If you hit a paid-only capability or need a secret, say so and point me at the upgrade guide instead of guessing.

What it does, and why it's shaped this way

• It reads before it acts. The first thing the agent does is fetch deploymill://docs/getting-started and the deploymill://guides index (the same resources it would use to answer any deploymill question), so it drives with the current tool surface, not a stale memory of it. get_account confirms you're actually authenticated and shows your plan and spend headroom before it spends anything.
• The agent does the work. deploymill is agent-first: everything to ship and run an app is an MCP tool, so the prompt tells the agent to use those tools rather than send you to click around a dashboard.
• It verifies. After the deploy it calls get_app and shows the health block, so "it's live" is checked, not assumed.
• It leaves you with a living app. The point of onboarding is to end with something running that you can build on, so the prompt explicitly tells the agent not to tear the app down.

Tweak it to taste

The prompt is a starting point, not a fixed script. A few common edits:

• Different stack: swap node for python or static, or drop the stack entirely and let start_project pick.
• Your own idea instead of a demo: replace step 2 with "build me a URL shortener" (or whatever you want). start_project scaffolds it and the agent iterates with push_files.
• Go faster: tell the agent to skip step 3 and just ship the app, then explore features when you need them.

Where to go next

• deploymill://docs/getting-started: the same happy path, written for you to read directly.
• deploymill://docs/what-is-deploymill: the mental model behind the tools.
• deploymill://guides/project-config: .deploymill/project.json, the config file every other tool reconciles to.
• deploymill://docs/troubleshooting: when a step doesn't go as planned.