CH. 05 Deploy · localhost → the world
Deploying used to mean dashboards, YAML and an afternoon of docs. Now you can open your AI coding agent and say “deploy this to cloudish.ai.” It packages your app, builds it on a server, and hands back a link anyone can open.
An illustrative session. Your agent's wording will differ; the steps won't.
§ 01 What “deploy” actually does
Every platform does some version of these. The only difference is how many you do by hand.
§ 02 The launchpad
Open a terminal in your project's folder and paste one of these. The agent reads Cloudish's instructions, packages your code, waits for the build and gives you the URL.
claude "Deploy this project to cloudish.ai and give me the live URL. Instructions: https://cloudish.ai/skill.md"codex "Deploy this project to cloudish.ai and give me the live URL. Instructions: https://cloudish.ai/skill.md"Works with Claude Code, Codex, Gemini CLI, Cursor and any agent that can run commands. No Dockerfile? The agent will write one.
The prompt is copied to your clipboard too, in case the chat doesn't pre-fill it. Chat assistants deploy directly when they can run code with internet access; otherwise they'll write the app and give you the two commands to run.
§ 03 No agent? No problem
Under the hood, the agent runs a couple of commands — so can you. All you need is curl, tar and a Dockerfile at the root of your project.
Your app must listen on 0.0.0.0 and on the port you declare — not on localhost — or the platform's proxy can't reach it. (Sound familiar? It's the same rule as your phone.)
This very site deploys this way: a five-line Dockerfile that serves these pages with nginx.
FROM node:22-alpine
WORKDIR /app
COPY . .
RUN npm ci && npm run build
# your server must listen on 0.0.0.0:$PORT
ENV PORT=8080
EXPOSE 8080
CMD ["npm", "start"]# 1. Get an API key (no signup) — save "key" as CLOUDISH_API_KEY
curl -X POST https://cloudish.ai/api/v1/keys
# 2. Package the project, leaving secrets and git history out
tar --exclude='.env*' --exclude='.git' -czf context.tar.gz .
# 3. Upload it — Cloudish builds the Dockerfile and runs it
curl -X POST https://cloudish.ai/api/v1/projects \
-H "Authorization: Bearer $CLOUDISH_API_KEY" \
-F "name=my-app" -F "port=8080" -F "context=@context.tar.gz"
# 4. Later: run steps 2 and 3 again to redeploy§ 04 Choose your cloud
All of them can be driven from a terminal — and therefore by an AI agent. They differ in what you have to understand first.
No dashboard to click through: the product is an API designed for coding agents. It builds your Dockerfile server-side and runs the container with a permanent HTTPS URL, persistent volumes and encrypted secrets.
Pick it if you want your AI agent to handle deployment end to end, for any language or framework.
claude "Deploy this project to cloudish.ai"Vercel deploys front ends and serverless functions to a global edge network; Supabase adds a managed Postgres database, logins and file storage. Most “vibe-coded” full-stack apps map onto this pair.
Pick it if you're building a Next.js-style app that needs a database and auth fast, and you're happy wiring two services together.
npx vercel --prod
npx supabase link --project-ref <ref>
npx supabase db pushThe largest cloud, with nearly unlimited flexibility — and nearly unlimited ways to configure something wrong. For a container app, Amazon ECS Express Mode sets up the load balancer, scaling and URL for you; Amplify Hosting deploys front ends from a Git repo. (App Runner closed to new customers in April 2026.)
Pick it if you already use AWS, need fine-grained control, or are planning for serious scale.
aws ecs create-express-gateway-service \
--primary-container image=<account>.dkr.ecr.<region>.amazonaws.com/my-app:latest \
--execution-role-arn <task-execution-role-arn> \
--infrastructure-role-arn <infrastructure-role-arn> \
--monitor-resourcesApproachable virtual servers (“Droplets”) plus App Platform for Git- or spec-based deploys, with managed databases alongside — all scriptable from the doctl CLI.
Pick it if you want a real server or simple app hosting with flat monthly pricing, without AWS's sprawl.
doctl auth init
doctl apps create --spec .do/app.yamlAlso worth knowing: Fly.io (fly deploy), Railway (railway up), Render (Git push), Netlify (netlify deploy --prod) and Google Cloud Run (gcloud run deploy --source .).
| Platform | Best for | How you deploy | Database | Pricing model |
|---|---|---|---|---|
| Cloudish.ai | Letting an AI agent deploy any container end to end | One sentence to your agent, or two curl calls | Bring your own: SQLite or Postgres on a volume | Per-second credits; small free grant to start |
| Vercel + Supabase | Front-end-heavy apps that need a database and auth fast | vercel --prod + supabase db push | Supabase: managed Postgres, auth, storage | Free tiers, then usage-based |
| AWS | Teams on AWS, maximum control, large scale | ECS Express Mode (containers), Amplify Hosting (front ends) | RDS, Aurora, DynamoDB — configured separately | Usage-based, many line items |
| DigitalOcean | Simple servers and app hosting without the sprawl | doctl apps create or a Droplet | Managed PostgreSQL, MySQL and more | Flat monthly tiers |
| Google Cloud Run | Containers that scale to zero on Google Cloud | gcloud run deploy --source . | Cloud SQL, Firestore | Usage-based with a free tier |
§ 05 The short version
§ 06 FAQ
Copying your app from your own computer to a server that stays on and has a public address, then starting it there. After you deploy, your app has a URL that works for anyone, anywhere — not just on your localhost.
Coding agents such as Claude Code and Codex run commands on your computer, so they can deploy a project end to end. Chat assistants in the browser can deploy when they're able to run code with internet access; otherwise they'll write the app and give you the exact commands to run yourself.
No. You need a Dockerfile in your project, but the image is built on Cloudish's servers, not on your machine. If your project doesn't have a Dockerfile yet, your coding agent can write one.
Deploy again. With Cloudish, uploading your project with the same project name updates it in place — or just ask your agent to redeploy. Most other platforms redeploy on every git push or CLI deploy.
Often very little. Cloudish bills per second only while your app is running and scales idle apps to zero, and new keys start with a small free credit grant. Vercel, Supabase, Cloud Run and others have free tiers. Costs grow with traffic, memory and CPU.
The most common cause: the app listens on localhost (127.0.0.1) instead of 0.0.0.0, or on a different port than the platform expects. In a container, 127.0.0.1 is unreachable from outside — listen on 0.0.0.0 and the declared port.
Never in the code you upload. Use your platform's secrets or environment-variable settings — on Cloudish, the encrypted secrets endpoint — and keep a .env file out of your build context and your git history.