Your agent handles the deployment from start to finish. It connects while, writes while.yaml, sends your code, and gives you your app’s address.

Prerequisites

An app in Node.js, Python or Go, or with its Dockerfile.
An AI agent: Claude Code, Codex, Cursor or Gemini CLI in a terminal, or a conversation with Claude, ChatGPT or Mistral Vibe.

Step 1: Give your agent these instructions

Open your agent in your app’s folder, and paste these instructions into it:

Instructions

Read while.sh/agents, then deploy my app with while.

Without an account, your agent deploys your app as a trial. With your account, it connects while’s MCP server, https://app.while.sh/mcp. On the first call, a page opens in your browser: there you authorize while, and you choose the app your agent works on.

Step 2: Your agent prepares the deployment

Your agent reads while’s catalog: what your app must follow, and the building blocks it can declare. It then writes while.yaml at the root of your app:

while.yaml

apiVersion: while/v1
project: mon-appli
services:
  - name: web
    port: 3000
    health: { path: /health }

Your app listens on the port given by the PORT variable, and answers on its health path as soon as it can serve. The other rules are in What your app must follow.

Step 3: Your agent sends the code

With a terminal, your agent sends the archive of your sources with a curl command, or pushes to the project’s git repository at while. In a conversation, it writes the files with while’s tools.

while then builds your app’s image and runs your tests. The new version gets its first visitors once it answers on its health path.

Step 4: Open your app

Your agent gives you your app’s address, for example https://mon-appli.while.cloud, over HTTPS.

When something breaks

A failing test stops the deployment, and your agent reads the error. A version that doesn’t answer on its health path never gets traffic, and the previous one stays live. Your agent reads the logs, fixes the problem and deploys again.

Your first deployment