Deployments

Python web apps (FastAPI, Flask, Django)

Deploy FastAPI, Flask and Django apps on ZevCloud. The start command your app needs, the port it should listen on, and the environment variables to carry over.

A FastAPI, Flask or Django project is not a server on its own. Your code defines an app object, and an app server runs it and listens for requests: uvicorn for FastAPI, gunicorn for Flask and Django.

If a Python service has no start command, the build falls back to running your main file, for example python main.py. That imports your app, defines it, and exits, because nothing started a server. The deploy succeeds, then the app is stopped for crashing repeatedly with nothing in its logs. The start command is the fix.

Settings at a glance

Framework	Start command	Port
FastAPI (`main.py`)	`uvicorn main:app --host 0.0.0.0 --port 8000`	`8000`
FastAPI (`app/main.py`)	`uvicorn app.main:app --host 0.0.0.0 --port 8000`	`8000`
Flask (`app.py`)	`gunicorn --bind 0.0.0.0:8000 app:app`	`8000`
Django (project `mysite`)	`gunicorn mysite.wsgi --bind 0.0.0.0:8000`	`8000`

main:app means “the app object in main.py”. If your app object or file has another name, change that part, for example uvicorn api.server:application --host 0.0.0.0 --port 8000.

Make sure the server is in your dependencies: add uvicorn (FastAPI) or gunicorn (Flask, Django) to requirements.txt or pyproject.toml.

Start command

When ZevCloud detects FastAPI, Flask or Django and you leave the start command empty, it fills one in for you:

On create, the recommended command is shown in the Start Command field, already filled in. It reads your repository to pick main.py or app/main.py for FastAPI, app.py or wsgi.py for Flask, and your project folder for Django. You can edit it before you create the service.
On deploy, a Python service with no start command gets the same command before it builds. It’s saved on the service, so you’ll see it in Settings → Build & Deploy, and the Setup check on the Overview tab says it was set for you.

ZevCloud never replaces a start command you set. Clear it and save to get the recommended one back on the next deploy.

The start command can’t use $PORT or other shell variables, so write the port number itself. If you change the service’s port, change the number in the start command too. A command ZevCloud filled in follows the port for you.

Port

ZevCloud routes requests to the port in Settings → Build & Deploy → Port (8000 for Python apps by default), and sets the PORT environment variable to the same number.

Your app has to listen on that port and on all interfaces (0.0.0.0). A server bound to 127.0.0.1 or localhost only accepts connections from inside its own container, so requests never reach it. The Setup check warns you when the port in your start command doesn’t match the service’s port.

Environment variables

Settings that differ between your laptop and production live in environment variables: database URLs, API keys, and allowed origins such as CORS_ORIGINS. Set them on the Variables tab, then redeploy. Variables reach your app on the next deploy.

A FastAPI app often reads its allowed origins like this:

import os
from fastapi.middleware.cors import CORSMiddleware

origins = os.getenv("CORS_ORIGINS", "").split(",")
app.add_middleware(CORSMiddleware, allow_origins=origins, allow_methods=["*"], allow_headers=["*"])

If CORS_ORIGINS isn’t set, the list is empty and every browser request from your frontend is refused. Set it to your frontend’s address, for example https://myapp.zevcloud.app,https://www.example.com.

ZevCloud reads your code for os.getenv("..."), os.environ["..."], os.environ.get("..."), pydantic BaseSettings fields and the keys in .env.example, and lists the ones that aren’t set:

on the create page, so you can fill them in before the first deploy;
in the Setup check on the service’s Overview tab, each with an Add button.

Variables your code reads with a default (os.getenv("DEBUG", "false")) are listed as optional. PORT is set for you.

Using an AI coding assistant?

Copy this prompt into your AI assistant

If an AI tool (Cursor, Claude, ChatGPT, Lovable, Bolt and similar) wrote your backend, it may run the app with python main.py or bind it to localhost. Give it this prompt:

This Python web app deploys to ZevCloud. Make sure it runs under an app server, not `python main.py`: for FastAPI use `uvicorn main:app --host 0.0.0.0 --port 8000`, for Flask `gunicorn --bind 0.0.0.0:8000 app:app`, for Django `gunicorn <project>.wsgi --bind 0.0.0.0:8000` (adjust the module path to where the app object actually lives). Add uvicorn or gunicorn to requirements.txt. Read every setting that differs between environments (database URL, secrets, CORS_ORIGINS and other allowed origins) from environment variables, list them all in a .env.example file with placeholder values, and tell me the exact start command to paste into the service settings.

Updated at, Monday, October 26, 2026