How Antideploy works, from one sentence to a live link.

The short answer

When you paste the sentence, your agent reads a page that tells it what to do. It logs in with one approval from you, creates your app, sends your project, and follows the deploy. Antideploy reads the code, builds it, runs it behind HTTPS, checks that it answers, and then keeps watching it.

The steps

1. Your agent reads the instructions

   The sentence points at agent.md, a procedure written for agents. It leads to a contract at /api/v1, which lists every endpoint, request shape, error code and limit in one unauthenticated request.

2. You approve one link

   The agent starts a device login and shows you a link and a code. You approve, and the agent saves a token on your machine. From then on it makes every call itself.

3. Antideploy reads your project

   It works out the language, framework, start command, port and what the app needs, and names the file each answer came from. It flags code that would deploy and then lose data, such as a SQLite file, uploads saved to disk or a scheduler inside the process. A project that matches a refused category stops here.

4. It builds a container

   Antideploy builds your project with standard buildpacks, or with your own Dockerfile if there is one at the root. A static site is built if it needs building and served as files.

5. It prepares the database

   If your code needs Postgres, one is created beside the app and DATABASE_URL is set. If you have a migration command, such as prisma migrate deploy, it runs before the new version goes live. A failed migration fails the deploy, so a broken schema never goes live.

6. It starts the app and checks it

   The app starts with your start command and the port in PORT. It has to answer a health check before the deploy counts as live. If it never starts listening, the deploy fails and says so.

7. It opens the app to the world

   The app gets its address, such as my-app.antideploy.app, with a certificate. Your agent tells you where it is live and on which account.

8. It runs a security check

   Right after the deploy goes live, Antideploy looks at what the running app shows the public: keys in the browser code, downloadable files such as .env, and CORS and header problems. It never holds the deploy up.

9. It keeps watching

   About every five minutes it checks that the app is running. You get an email when an app goes down and another when it is back. Your agent can read the logs, metrics and health at any time.

Where each step can stop

Every step either passes or stops with a plain reason.

• Analysis. The project is not recognised, or it matches a refused category.
• Build. A dependency cannot be installed, or the code does not compile.
• Migration. The schema change fails, so the new version is not released.
• Start. The app never listens on its port, or it crashes on boot.
• Security check. It reports findings, but it never fails the deploy.