Vercel Alternatives: Heroku, Railway, Render
Short answer: If your app is a Next.js front end, Vercel is hard to beat. If it is a full app with a database, background jobs or WebSockets, you pay a platform to hide a server from you, and the bill grows with usage, seats and bandwidth. The alternatives are Railway, Render and Fly.io (usage-based or per-instance platforms), Heroku (which went into maintenance mode in February 2026), or your own server with git-push deploys, such as a NameOcean App Server: 3 vCPU and 3 GB of memory for $5 a month with your database on the same machine. Prices below were read from each vendor's own pricing page on 20 November 2026.
Why people look for an alternative
- Usage billing is hard to predict. Vercel charges for active CPU time, memory, invocations and data transfer; Railway and Fly.io charge per second of CPU and memory. A traffic spike or a slow query becomes a bigger bill.
- Per-seat pricing. Vercel Pro is $20 per developer per month, so a team of four pays $80 before any usage.
- Serverless limits. A Vercel function runs for at most 300 seconds by default (800 on Pro). Long jobs, queues and WebSockets need something that stays on.
- Small instances. The cheapest always-on plans on Heroku and Render have 512 MB of memory. A Django or Next.js app with a few workers outgrows that quickly.
- Heroku's future. On 6 February 2026 Heroku announced a move to a "sustaining engineering model" focused on stability and security, meaning no new features. Existing apps keep running, but many teams have started planning a move.
Let your AI deploy it
Connect Claude Code to NameOcean and ask it to create a server, deploy your app and point your domain. App servers from $5/month, billed by the hour.
claude mcp add --transport http nameocean https://nameocean.net/mcp
Cursor, VS Code and other clients →
The options at a glance
The example is a small always-on web app with a Postgres database, using each platform's cheapest plan that does not sleep.
| Cheapest always-on app | Postgres | Free tier | Shell access | |
|---|---|---|---|---|
| Vercel | Pro $20/month per developer, includes $20 of usage; then active CPU from $0.128/hour | Not its own: a Marketplace provider (Neon and others) | Hobby: personal, non-commercial use only | No |
| Heroku | Basic dyno $7/month, 512 MB | Essential-0 $5/month, 1 GB | None | heroku run bash (a separate one-off dyno) |
| Railway | Hobby $5/month including $5 of usage; then about $20 per vCPU-month and $10 per GB-month | Same usage rates | 30-day trial, then $1/month of credit; free services sleep | railway ssh |
| Render | Starter $7/month, 0.5 CPU, 512 MB | From $6/month (256 MB memory) | Free web service that spins down after 15 minutes idle; free Postgres expires after 30 days | SSH on paid services |
| Fly.io | shared-cpu-1x, 512 MB: $3.69/month, usage-based | Managed Postgres from $38/month | No free allowance for new accounts | fly ssh console |
| NameOcean App Server | Starter $5/month: 3 vCPU, 3 GB, 30 GB NVMe, 20 TB traffic, billed by the hour | Runs on the same server at no extra cost | None | Root SSH |
Bandwidth is the other line to watch: Render includes 5 GB on Hobby and 25 GB on Pro and then charges $0.15/GB; Vercel lists $0.15/GB after 1 TB unless your team is on its newer flat-rate CDN; Railway charges $0.05/GB of egress and Fly.io $0.02/GB in Europe and North America.
Vercel alternative: when to switch and when not to
Stay on Vercel for a static or Next.js front end with light API routes: preview deployments for every pull request, the global CDN and image optimisation are what you pay for, and you would rebuild them yourself elsewhere.
Move the back end off Vercel when you need a process that keeps running (queues, cron, WebSockets, a Discord bot), a database next to the app, or a bill that does not depend on CPU seconds. A common split is the front end on Vercel and the API plus database on a server, with the front end calling it on its own subdomain.
Heroku alternative
Heroku's model (git push, a Procfile, config vars, add-ons) is the one most alternatives copy, so moving is mostly mechanical:
- Render and Railway are the closest in feel: connect the GitHub repository, set the start command and environment variables, add a database.
- Fly.io deploys a Docker image to machines in the regions you choose. It suits apps that need to run close to users in several places.
- Your own server keeps the git push but replaces dynos with one bigger machine. The
web:line of yourProcfilebecomes a supervisor program, config vars become environment variables, andpg_dumpfrom Heroku Postgres loads into Postgres on the server.
Railway and Render alternatives
Railway and Render are good platforms. People leave them for three reasons: the bill on a usage-based plan (Railway) moved more than expected, the 512 MB instance (Render Starter) ran out of memory and the next size up costs $25, or they needed something the platform does not allow, such as a custom system package, a long-running process or root access.
Your own server, without the server work
A plain VPS is the cheapest way to run an app, but you configure nginx, certificates, deploys and backups yourself. A NameOcean App Server comes with those done:
- git push deploys. Each server has a git remote. After a push, the server checks out your code and runs
.nameocean/deploy.shfrom your repository, where your install and build steps go. - HTTPS and domains. Every server gets a free
https://<id>.nameocean.orgaddress; your own domains get a certificate automatically. - Daily backups, kept for 7 days and restorable from the dashboard.
- A whole Linux server. Rocky Linux, AlmaLinux, Debian or Ubuntu, root SSH, any language, Postgres, Redis or a queue on the same machine.
- An API, a CLI and an MCP server. Claude Code, Cursor and other AI tools can create the server, deploy, read logs and set environment variables for you.
In our test, Claude Code created a Starter server and had a Flask app live on HTTPS in 2 minutes 27 seconds; the next change took 65 seconds.
What is different
- Location. Servers run in Helsinki, Finland today, close to Europe and the Middle East. Need your users' region? Tell us: new locations are opened on request.
- Scaling is one click, not automatic. You choose a size (3 to 10 vCPU) and move to a bigger plan with one click in the dashboard, one API call or one sentence to your AI (
resize_server). Nothing scales up behind your back, so the bill never jumps on its own. - No managed database. Postgres on your server is yours to run; the daily server backup includes it, but there is no point-in-time recovery or read replica unless you set them up.
- No preview deployments per branch yet, and no free tier.
Moving an app in practice
- Create the server (dashboard,
nameocean servers create, or ask your AI). - Add a
.nameocean/deploy.shthat installs dependencies and runs migrations, and a supervisor program with the same command as yourProcfile'sweb:line, listening on127.0.0.1:8000. The git push guide has examples for Django, Flask, Node.js, Next.js and Go. - Copy your environment variables (
nameocean servers env set), thengit pushto the server's remote. - Restore the database:
pg_dumpon the old platform,pg_restoreon the server. - Test on the
nameocean.orgaddress, then bind your domain and switch DNS.
With the MCP server connected, the same steps are one request: "Move this Heroku app to a new NameOcean server: read the Procfile and app.json, set up the deploy files, copy the config vars I paste, and tell me when it answers on the test URL."
FAQ
What is the cheapest Vercel alternative for a full-stack app?
For an always-on app with a database: a server that runs both. Fly.io's smallest machine is cheaper than $5 but its managed Postgres starts at $38; Heroku and Render need two products (app plus database, $12 to $13 a month) for 512 MB of memory. A $5 NameOcean Starter server has 3 GB and runs the database too.
Is Heroku shutting down?
No. In February 2026 Heroku moved to "sustaining engineering": it keeps the platform running and secure but stops adding features. Card-paying customers were told pricing and service stay the same.
Can I deploy Next.js on my own server?
Yes: npm ci && npm run build in deploy.sh and next start as the supervisor program. You lose Vercel's edge network and image optimisation unless you add a CDN, and you keep server-side rendering, API routes and middleware.
Do I need to know Linux?
Less than with a plain VPS: HTTPS, nginx, the firewall and backups are set up for you, and an AI client with the MCP server can do the rest. It still helps to read the logs it shows you.
Sources
Let your AI deploy it
Connect Claude Code to NameOcean and ask it to create a server, deploy your app and point your domain. App servers from $5/month, billed by the hour.
claude mcp add --transport http nameocean https://nameocean.net/mcp
Cursor, VS Code and other clients →
Read in other languages: