Hosting · checked 2026-10-20

Can I vibecode Render?

KINDA· weekend project

A personal deploy server is a real weekend project, not a sitting: webhook, build, health-checked container swap, Caddy routing, env vars, a Postgres container with backups and cron are all ordinary pieces, and an agent can wire them on a single VPS. What stays behind is what the monthly bill pays for: multiple regions, autoscaling, managed Postgres with point-in-time recovery, DDoS filtering and someone else being paged at 3 a.m. Coolify, Dokploy and CapRover already do the self-hosted version, so writing your own is mostly worth it if you want something smaller than they are.

$7/moprice

$84you'd save a year

weekendbuild time

—people replaced it

Before you start

What you need on hand.

DifficultyAdvanced · Real infrastructure or OAuth work.

TimeA weekend; the server and DNS setup is the slow part

Claude Code, Codex or CursorAny of them can run the prompt.
A VPS with Docker and CaddyAny small Linux server works; budget a few dollars a month.
Node.js 22better-sqlite3 compiles a native module, so build tools must be installed.
A domain and DNS accessEach app needs an A record pointing at the server before Caddy can get a certificate.
Your Render settingsA generated Blueprint, the environment variables of each service and a pg_dump of each database.

Before you buildRender's Hobby workspace costs nothing and a Free web service is fine for a demo. It sleeps after 15 minutes and the free Postgres expires after 30 days, so it is a trial, not a home.

1. Install one agent: Claude Code, Codex, Cursor.
2. Make an empty folder and open it in the agent.
3. Paste the prompt below, press enter, and answer its questions as it builds.

The prompt

Build me a personal deploy server to replace Render for my side projects. Requirements:

- Node.js 22 + TypeScript on one Linux VPS with Docker and Caddy already installed. A Fastify
admin UI (server-rendered, a little vanilla JS) listens on 127.0.0.1:4900; I reach it through
an SSH tunnel. A second port, 4901, only accepts GitHub webhooks.
- Storage: better-sqlite3 in ./data/deployd.db: apps (name, repo URL, branch, port, health path,
domains), env vars encrypted with AES-256-GCM using a key from .env, deploys (commit, status,
log) and cron jobs. Source and build logs live under ./data/src and ./data/logs.
- Deploy: on a push webhook (HMAC-SHA256 check of X-Hub-Signature-256) or `deployd deploy <app>`,
`git clone --depth 1` the commit, build with its Dockerfile (or `pack build` with the Paketo
builder when there is none), and tag the image with the commit SHA.
- Release: start the new container on a private Docker network, wait up to 60 seconds for the
health path to return 200, then point Caddy at it through its admin API (localhost:2019) and
stop the old one. A failed health check keeps the old container and marks the deploy failed.
`deployd rollback <app>` re-points to the previous image; keep the last three images.
- Domains: each app lists its domains; Caddy gets TLS certificates on its own once DNS points here.
- Postgres: one postgres:17 container, a database and user per app injected as DATABASE_URL, and a
nightly pg_dump to ./data/backups with 7 days of retention (`deployd db restore <app> <file>`).
- Cron jobs: a schedule and a command per app, run with `docker run --rm` from the app's image
using node-cron; the last 20 runs and their output are shown in the UI.
- Single user, one admin token in .env. No accounts, no telemetry, local-first. Outbound calls:
GitHub clones, image pulls and certificate issuance only.
- Out of scope: autoscaling, several servers, regions, preview environments, managed Redis.
- README: VPS setup, the DNS records, a firewall allowing only 22, 80 and 443, and two plain
warnings: one server means downtime on every reboot, and a compromised host exposes every env var.

Opens your agent with the prompt filled in, you press enter. Or copy it and paste it into an empty folder.

Share on X

Max MiskowSponsoredAI agents, automations and MVPs, built for your company.Visit site ›

Then keep going

Next prompts, same folder.

The first prompt builds the core. Run these one at a time in the same folder to grow it into your own Render.

Extend the deploy server in this folder. Requirements:

- On a pull_request webhook (opened, synchronize), build the branch and run it as a separate
  container at pr-<number>.<app domain>, with Caddy routing added through its admin API.
- Give each preview its own database cloned from the app's database with `createdb --template`,
  and inject that DATABASE_URL; skip the clone for apps without Postgres.
- Remove the container, route and database when the pull request closes, and after 7 days of no pushes.
- Post the preview URL as a comment on the pull request with a GitHub token from .env.
- Update the README with the wildcard DNS record this needs.

Add off-server backups to the deploy server in this folder. Requirements:

- After each nightly pg_dump, encrypt the file with `age` using the public key in .env and upload
  it with the S3 API (`@aws-sdk/client-s3`) to a bucket from .env, keeping 30 daily copies.
- `deployd backup verify` downloads the newest file, decrypts it into a throwaway postgres
  container, runs `SELECT count(*)` against each table and reports pass or fail.
- Run the verify step weekly and show the result and the age of the last good backup on the UI home page.
- Back up ./data/deployd.db and the Caddy data directory in the same job.
- Update the README with a restore walkthrough from an empty server.

Add a Render importer to the deploy server in this folder. Requirements:

- `deployd import-render render.yaml` reads a Render Blueprint (`yaml`) and creates apps for its
  web services, workers and cron jobs: repo, branch, build and start commands, health check path,
  and domains.
- Variables with sync: false are listed and prompted for; fromDatabase and fromService
  references are mapped to this server's own DATABASE_URL and internal hostnames.
- Print a table of what was created and what was skipped (static sites, disks, regions, plans)
  with the reason, and write nothing until I confirm.
- Update the README with how to generate the Blueprint in Render and how to move a database with pg_dump.

What you lose

Managed Postgres with point-in-time recovery and automatic failover
Regions around the world, a CDN and DDoS filtering
Autoscaling and zero-downtime deploys across several instances
Preview environments for every pull request
Someone else keeping the servers patched and on call

Why people still pay

A small team gets deploys, databases, TLS and logs without owning a server, and the first outage on a self-managed box usually costs more than the $7 it saved. Render also lets you grow from one service to a full stack with private networking, workers and cron on the same bill.

Moatinfrastructure scale execution polish

Do the math

Render or your own copy?

Over

Your usage

Count my time12h ×$/h

Render, 1 yr$84

Your own copy, 1 yr$129running + build session

Paying is cheaper by$45never pays for itself

Estimate · A VPS with a couple of GB of RAM runs several small apps and a Postgres container for a few dollars a month; add storage for backups if you ship them off-box. The build is token usage on your agent plan.

Leaving Render

Take your data, then cancel.

1 · Take your data out of Render

1Services: in the Dashboard select your services and click Generate Blueprint. You get a render.yaml with each service's type, plan, build and start commands, but environment variable values are left out (set to sync: false), so copy them from each service's Environment tab.
2Postgres: run pg_dump from your machine against the database's external URL (pg_dumpall for several databases). Paid databases can also make a logical export from the Recovery page, downloadable as .dir.tar.gz and kept for seven days; Free databases have no backups.
3Disk contents have no one-click export in the pages we read: copy them out over SSH from the service shell, which Free instances don't have.
4Key Value usually holds cache data, and Free instances are in-memory only, so there is often nothing to carry over.

2 · Cancel Render

1Delete or suspend each web service, worker, cron job, database and Key Value instance from its Settings page. Compute is billed prorated to the second, so charges stop when the resource is gone.
2To stop the workspace fee, open the workspace dropdown, choose Billing, and pick the Hobby plan; the change applies immediately.
3Leaving a Scale workspace for Pro or Hobby first needs the extra workspaces deleted so only the primary one remains.
4Check the Billing page afterwards for usage-based charges that land at month end, such as bandwidth and build minutes.

A community thread reports that a Postgres database can't be downgraded in place: take a dump, create a smaller instance and restore into it.
Free Postgres is deleted 30 days after creation with a 14-day grace period, so dump anything you care about before then.
Lower your DNS TTL a day ahead and keep the Render service running until the new server answers, then switch the record.
A service that writes files to a persistent disk loses them when you delete it, so copy them out first.

Checked 2026-10-20 against Render's docs: 1, 2, 3, 4, 5, 6

The escape hatch

Don't feel like building it? These are already free.

CoolifyOpen sourceApache-2.0 self-hosted platform for apps, databases and services on your own server, with Git deploys, automatic TLS and backups.★ 63k · active Oct 2026 · one-click install
Web, LinuxVisit ↗DokployOpen sourceSelf-hosted deployment platform for Docker apps and databases with Git deploys; Apache-2.0 with a proprietary directory in the repo.★ 38k · active Oct 2026 · one-click install
Web, LinuxVisit ↗DokkuOpen sourceMIT mini-Heroku that runs on one server: git push to deploy, buildpacks or Dockerfiles, plugins for Postgres and TLS.★ 32k · active Oct 2026 · self-host, real ops
LinuxVisit ↗

All 5 free alternatives to Render ›

The numbers

Render pricing

Plan	Monthly	Annual, per month	What you get
Hobby workspace	$0/workspace	—	Up to 25 services, 5 GB of bandwidth, 500 build pipeline minutes, single-service previews, custom domains, firewall and DDoS mitigation, chat support. Compute is billed on top.
Pro workspace	$25/workspace	—	Unlimited seats and services, 25 GB of bandwidth, 1,000 build minutes, full-stack previews, horizontal autoscaling, isolated environments, audit logs. Compute is billed on top.
Scale workspace	$499/workspace	—	Everything in Pro plus multiple workspaces, 1 TB of bandwidth, HIPAA-compliant workspaces, SAML SSO and SCIM, advanced roles. Compute is billed on top.
Enterprise	custom	—	Contractual uptime SLAs, a dedicated technical account team, a support Slack channel.
Web service: Starter	$7	—	512 MB RAM, 0.5 CPU, always on, billed prorated to the second.
Postgres: Basic	$6	—	256 MB RAM, 1 GB storage included, point-in-time recovery, logical exports.

Free tier: Free web services sleep after 15 minutes idle and share 750 instance hours per workspace a month, with ephemeral disk and no shell. One free Postgres per workspace, 1 GB, deleted 30 days after creation (plus a 14-day grace period), no backups. Free Key Value is in-memory only.
Billing: Workspace fee monthly, compute prorated to the second, usage billed at month end. The pricing page shows monthly figures only; no annual discount is mentioned. Cards through Stripe.
Hidden costs: Bandwidth past the included amount is $0.15 per GB, build minutes past the included amount are $5 per 1,000, persistent disks are $0.25 per GB a month, extra custom domains $0.25 each, and dedicated IPs $100 a month. A small app is easily a Starter instance plus a $6 database, so $13+ before the workspace plan.
Source: render.com

Also on the list

Vercel$20/mo · HostingKINDA

Netlify$9/mo · HostingKINDA

Heroku$7/mo · HostingKINDA

Questions

Render, briefly

Not answered here? Tell us and we'll add it.

Ask a question ›

The Hobby workspace is $0, and Free compute plans cover web services, Postgres and Key Value with tight limits.

Kinda. The core is a weekend build with the prompt on this page, but real gaps remain: Managed Postgres with point-in-time recovery and automatic failover, Regions around the world, a CDN and DDoS filtering.

About $7 a month on the Hobby workspace + Starter instance plan, $84 a year (checked 2026-10-20).

Managed Postgres with point-in-time recovery and automatic failover; Regions around the world, a CDN and DDoS filtering; Autoscaling and zero-downtime deploys across several instances; Preview environments for every pull request; Someone else keeping the servers patched and on call. If any of those matter to you, keep paying.