No Dockerfile
The platform detects Django from requirements.txt / pyproject.toml, picks the build and start commands, and lists the environment variables your code reads.
Django deploy guide · Agent Lab Deploy
Deploy Django
in one sentence.
Upload your Django project as it is — no Dockerfile — or tell Claude Code, Codex or Cursor "deploy this". Agent Lab builds it into a web service, gives every version a preview URL with rollback, and checks uploads for leaked keys. Apps and databases run on Google Cloud in Taiwan.
> deploy this Django app
sites_createmy-django
sites_deployDjango detected
sites_deploy_statusbuilt · secret scan passed
Live at my-django--acme.hosting.agentlab.run
Why deploy Django on Agent Lab
Django online without running a server yourself
Python project, built from source with no Dockerfile, deployed as a web service.
Preview, go live, roll back
Every upload is a new version with its own URL. Look first, then go live; switch back to any version in one click.
Leaked keys caught
Uploads are checked for API keys and credentials before they can go live. Secrets go in the console, never in the code.
PostgreSQL in one click
Create a database and connect it; the connection is written into the app's environment and the password is never shown.
Your options
Django on Agent Lab, or on your own VPS
Scroll the table sideways.
| What you care about | Agent Lab Deploy | Your own VPS |
|---|---|---|
| Setup | Upload or say one sentence; no Dockerfile | SSH, a runtime, a process manager, a reverse proxy |
| HTTPS and domains | Automatic, custom domains on paid plans | Certbot and DNS by hand |
| Preview and rollback | Every version has its own URL; roll back in one click | Write your own scripts |
| Before going live | Uploads checked for leaked keys | Nothing unless you set it up |
| Database | PostgreSQL in one click, connection written into the app | Install and back it up yourself |
| Cost to start | Free plan: NT$0, one size-S app (sleeps when idle) | A VPS at roughly NT$200–600 a month, plus your time |
Steps
Deploy Django in four steps
Prepare
Keep requirements.txt / pyproject.toml at the top of the project and commit the lockfile. Leave out node_modules, .git, dist and .env.
Upload or ask
Drop the folder on a project's Deploy tab, push from GitHub, or ask your AI tool through Agent Lab MCP.
Build and check
The platform builds the version and checks it for leaked keys. Enter any secrets it lists in Settings.
Preview and go live
The first version goes live once it builds. Later ones get a preview URL; go live when you're happy.
Setup
What Agent Lab runs for Django
Serve static files with WhiteNoise and run collectstatic at build time; the filesystem is read-only at run time. Read DATABASE_URL with dj-database-url and connect an Agent Lab PostgreSQL. Don't run migrate on every start; have your AI tool run it once before going live.
Build and start
# Agent Lab 偵測到 Django 後自動產生 WSGI 伺服器指令
# Agent Lab detects Django and generates the WSGI server command
pip install -r requirements.txt # 也支援 Poetry、uv
# 啟動 WSGI app，監聽 $PORT（預設 8080）
# Python 版本：.python-version → pyproject → 3.12SECRET_KEYgenerate it in the console; keep it out of settings.py.ALLOWED_HOSTSmust include your Agent Lab address and any custom domain.DEBUGset False in production.DATABASE_URLwritten for you when you connect an Agent Lab PostgreSQL.
Or ask your AI tool
Read https://app.agentlab.cresclab.com/agent/deploy.md and deploy this Django project to Agent Lab. Show me the build plan and the environment variables it needs first, tell me where to enter any secrets, and wait for my OK before it goes live.FAQ
Deploying Django: questions
Do I need a Dockerfile to deploy Django?
No. Upload the Django project with requirements.txt / pyproject.toml at the top and Agent Lab detects the framework, builds it and starts it. A Dockerfile at the top, if there is one, is used instead. Docs →
Do I need to configure gunicorn for Django?
No. The platform detects Django and generates the WSGI server command on PORT. To customize it (worker count, for example), set a Start command in the build settings or add a web: line in a Procfile.
Where do API keys and other secrets go?
In the project's Settings, under Secrets; the code reads them as environment variables. Random values such as session secrets can be generated in the console without anybody seeing them. A key found in an upload stops it before it goes live.
Can I deploy Django from Claude Code or Codex?
Yes. Connect Agent Lab MCP once (one command, sign in in the browser, no API key), then say "deploy this Django app". The tool creates the project, uploads, waits for the build and asks before going live. Connect MCP →
Is there a free way to host Django?
The Free plan is NT$0: one size-S app that sleeps after 15 idle minutes, one size-S database, three static web pages, 100 build minutes and 5 GB of traffic a month, pausing at the limit instead of charging. Pro (US$29 a month before tax) keeps apps running and adds custom domains. Pricing →