AI hosting: your AI deploys your app with a single token

What AI hosting means

AI hosting is cloud hosting designed to be operated by an AI assistant, not only by a person clicking through screens. Instead of you learning commands and settings, the AI reads a public technical manual, receives a credential for your account and talks directly to the platform API.

In practice, your AI gets the same hands you would have in the dashboard to deploy: create a project, create a database, set variables, ship a new version and read logs. You still decide the plan, the payments and what goes live.

The manual the AI reads lives at https://devoracloud.com/llms.txt. It is written for language models, with the steps, the limits and the most common errors. That is why it works with any assistant that can read a URL and make HTTP calls: ChatGPT, Claude, Gemini, Cursor, GitHub Copilot and others.

Step by step: from sign-up to a live site

• Create your account at devoracloud.com/register and confirm your e-mail. Every new account gets a 7-day free trial on the Hobby plan, no card required.
• In the dashboard sidebar, open “Token de IA e API” (AI token & API) and click “Gerar token de IA” (Generate AI token).
• Copy the text shown. It already contains the token (starting with dc_live_) and the instructions for the AI.
• Paste it into ChatGPT, Claude, Gemini, Cursor or Copilot, together with your code or a description of the app.
• Follow the conversation: the AI reports each step and, at the end, hands you the site address.

You do not need to create the project first, install Git or write a Dockerfile. If you have no account yet, the AI sends you the sign-up link and waits for the token.

What the AI does for you

With the token, the AI follows the llms.txt playbook. A typical deploy goes through these steps:

• Reads llms.txt and checks access by listing the account projects and plan.
• Creates the project in ZIP upload mode.
• Creates the PostgreSQL (or Redis) database, if the app needs one, before the first deploy. The connection reaches the app as DATABASE_URL, without you typing a password.
• Creates the project’s private S3 bucket when the app stores images, PDFs or videos.
• Zips the code without node_modules, .env or build folders, and uploads it.
• Starts the deploy and follows the status until READY, reading the logs if something fails.
• Gives you the address https://your-project.apps.devoracloud.com and, if you want, sets up your custom domain.

If the build breaks, the AI reads the deploy diagnosis, fixes what was missing (a dependency, the start script, the port) and tries again. A failed deploy never replaces the version already running.

After launch: the AI diagnoses too

The token is not only for the first deploy. When the site gets slow, restarts on its own or runs out of space, you can ask the same AI what is going on. It checks CPU and memory metrics for recent hours, the restart count, the application logs and S3 usage before giving an opinion.

With that data, the answer stops being a guess. The AI can show, for example, that memory is at the limit and the logs say “out of memory”, or that CPU and memory are fine and the bottleneck is a slow database query. We wrote a separate guide on exactly that: how to tell whether you lack CPU, memory, or the code is the problem.

Deploying yourself vs deploying with AI

Security: treat the token like a password

The token belongs to one account and grants API access to that account. A few simple rules:

• Paste the token only into assistants you trust and into private conversations.
• The AI must not echo the full token or put it in code, commits or logs.
• The AI does not buy plans, replicas or storage packs for you: payments stay in the dashboard, by card (or Pix in BRL for customers in Brazil).
• If you suspect a leak, revoke the token under “Token de IA e API”. Anyone using it loses access immediately.

Today applications run in the USA. Brazil has Devora’s own CDN cache, which speeds up delivery to Brazilian visitors but does not run applications.

Pricing

You start with a 7-day free trial, no card. Then you pick a plan, billed monthly in US dollars:

• Hobby — US$ 5: 0.5 vCPU, 512 MB RAM, 1 database and 1 GB of S3.
• Pro — US$ 20: 2 vCPU, 1 GB RAM, 2 databases, 5 GB of S3 and autoscaling.
• Enterprise — US$ 59: 4 vCPU, 8 GB RAM, 4 databases, 25 GB of S3 and autoscaling.

Every plan includes a stopped failover replica on another server, which takes over if the primary instance fails.

Frequently asked questions