Introducing deplo.ai — code goes in, production URL comes out

Every developer knows the moment: the app works locally, and now it needs to exist on the internet. What follows is rarely engineering. It's choosing a host for the frontend, another for the backend, learning two dashboards, guessing at build commands, copying environment variables between tabs, discovering CORS, and — an hour later — realizing the API URL baked into the frontend points at localhost.

deplo.ai collapses all of that into one action. Connect a GitHub repository, click Deploy (or type deplo), and a few minutes later you have two production URLs: the frontend live on Vercel, the backend live on Render, wired to each other correctly.

The unusual part: we own nothing

Most deployment platforms simplify hosting by becoming your host — and that convenience compounds into lock-in. deplo.ai makes the opposite bet. Deployments land in accounts you own: the frontend appears in your Vercel dashboard like any project you created by hand, the backend runs as a normal web service in your Render workspace. Your URLs, your logs, your billing relationship, your custom domains.

The consequence we're proudest of: if deplo.ai vanished tomorrow, every app ever deployed with it would keep running, untouched. A platform should have to earn your continued use — not hold your infrastructure hostage.

What actually happens when you deploy

Under the single Deploy action is a pipeline that treats your repository as the source of truth rather than asking you to describe it:

• Detection is deterministic first. A weighted-signature engine reads your repo tree — dependencies, config files, scripts, structure — and identifies the stack: Next.js in client/, Express in server/, a Vite app at the root, a FastAPI service, a Go binary. Monorepos and workspaces included.
• Commands come from your code. Node backends run the scripts in your own package.json; Python apps get their start command from the actual app module we find; Go builds follow your layout. Nothing is guessed that can be read.
• Env vars are found, not requested blindly. The scanner reads your source for the variables the code actually consumes — not just .env.example — marks what's required, and asks for only that.
• Wiring is automatic. The backend deploys first; its live URL is injected into the frontend build under the keys your code really uses (NEXT_PUBLIC_API_URL, VITE_API_URL, …), and the backend receives your frontend origin for CORS. The localhost-in-production bug is structurally impossible.
• Pushes redeploy. After the first successful deployment, a webhook redeploys the branch on every push, reusing the proven configuration.

AI on a leash

There's AI in deplo.ai, but it's deliberately bounded: at most two model calls per deployment. One when deterministic detection isn't confident about an unusual repo layout; one when a deployment fails — producing a diagnosis with the likely cause, the affected file, the fix, and a prompt you can paste straight into Cursor, Claude Code, or Copilot. A failed deploy shouldn't be a wall of logs; it should be an explanation.

A wrong AI answer can never override a working deterministic result. We think that's the only honest way to put a model inside infrastructure tooling.

The terminal is the product too

Everything above works without opening the dashboard at all:

$ npx @deplo.ai/cli login
$ cd your-project
$ npx @deplo.ai/cli
→ repo detected   github.com/you/your-app (main)
→ stack detected  frontend=Next.js · backend=Express
✓ backend live    https://deplo-your-app-api.onrender.com
✓ frontend live   https://deplo-your-app-web.vercel.app
● Deployed in 2:41

The CLI is a thin client over the same API as the dashboard — same deployments, same monitoring, same account. It runs in CI with a token and exit codes, so deploy can gate a pipeline. The CLI docs cover every command.

After the deploy

Shipping is the start, not the end. Every project gets uptime and latency monitoring, live CPU and memory from your Render service, and optional cookieless product analytics — views, visitors, top pages — via a one-line snippet. When something breaks, the failure email tells you what, where, and how to fix it.

Free while in beta

deplo.ai is free during beta. Your apps run on your providers' free tiers unless you choose otherwise, so the realistic cost of shipping a full-stack side project today is zero.

Start at the quickstart, or just run npx @deplo.ai/cli login. Code goes in. Production URL comes out.