A Node.js, Next.js or Python app you wrote yourself can be moved off Zeabur in five steps: get your source code and the real values of your environment variables, export your database and uploaded files, deploy on the new host and test it on its default address, switch your domain's DNS, and only then walk through sign-in, forms and webhooks one last time. Zeabur recommends finishing the move by November 30, and all services stop on December 31, when service data is deleted. Starting now still leaves enough time.
This is the hands-on version. We use RoamerHost's Vibe Coding Hosting as the destination, but most of the commands and gotchas apply wherever you go. What happened at Zeabur, how refunds and claims work, and how to move n8n or WordPress are covered in our Zeabur alternative and migration guide, so we won't repeat them here.
Zeabur's timeline, CLI commands and platform behavior come from Zeabur's migration guide of October 20 and its announcement of October 22, checked on November 11, 2026. If Zeabur updates them, its official pages take precedence.
Before you start: list what your project relies on
Most migration mistakes come from things Zeabur did for you automatically. Zeabur's own guide spells them out: generated *.zeabur.app addresses stop resolving, the .zeabur.internal names your services used to reach each other stop working, and images built from Git live in Zeabur's own registry, which goes away with the platform. Deploy on push, deployment history, one-click rollback and Zeabur's automatic database backups all have to be rebuilt somewhere else.
Log in to the Zeabur CLI with the account that owns the project and record each service's settings. These commands run on your own computer:
npx zeabur@latest auth login
npx zeabur@latest project export --id <projectID>
npx zeabur@latest service list --project-id <projectID>
npx zeabur@latest --json service get --id <serviceID>
npx zeabur@latest --json variable list --id <serviceID>
npx zeabur@latest domain list --id <serviceID>
The project ID is the 24 characters after project- in the dashboard URL. service get shows the source repo, branch, root directory and any custom build or start commands, which you'll need to recreate. The output of project export and variable list contains passwords and keys. Zeabur's guide says to treat it as secret: keep it only in files with 600 permissions on your own machine, never paste it into a chat, and never commit it.
If you work with Claude Code or Codex, Zeabur also wrote a migration reference for coding agents. Ask your agent to read https://zeabur.com/blogs/migrate-from-zeabur.md and take inventory with you. Let it use the CLI you've already logged into, and don't hand it your Zeabur token.
Step one: get your code back
If the service was deployed from GitHub, the code is already in your repo. What you need to copy are Zeabur's settings: whether the root directory is a subfolder, and whether a zbpack.json file or ZBPACK_ variables changed the build and start commands. Zeabur detects projects with its open-source zbpack tool, and none of that configuration follows you automatically.
If you deployed by uploading code and no longer have the source, Zeabur's guide suggests exporting files from the running container: run tar inside it with service exec, pipe the output through base64 and decode it locally. The guide has the full command. Be aware that what comes back may be built output rather than source. A static site can be served as is, but an app with a backend will be painful to change later, so recover the real source if you possibly can.
Once you have the code, search the repo for zeabur.app and zeabur.internal. Hard-coded URLs, OAuth callback URLs and webhook settings for payments or LINE often still point at the old addresses.
Step two: get the real values of your environment variables
Environment variables are the easiest thing to miss, for two reasons. Zeabur variables can reference other variables with ${VAR}, and both variable list and project export return the unexpanded form, such as ${PASSWORD}. And apps often connect to their database through variables shared by another service, like POSTGRES_CONNECTION_STRING or REDIS_URI. Those sit under readonlyVariables and don't show up in the service's own list. The safest approach is to search your code for every variable it actually reads and confirm each one has a value in the new environment.
Variables Zeabur injected for you need replacing too. Public URLs like ZEABUR_WEB_URL and service addresses like <SERVICE>_HOST won't exist anymore. Check the port as well: Zeabur's Git services listen on 8080 unless another port was detected. Vibe Coding Hosting expects your app to listen on 0.0.0.0 and the PORT environment variable, so if 8080 or 3000 is hard-coded, read PORT instead.
One more thing is not optional. Both of Zeabur's security incidents this year exposed environment variables, so every API key, database password and third-party token that lived on Zeabur and is still valid should be regenerated before it goes into the new environment. On Vibe Coding Hosting you set variables on the project page or through your AI; they're stored encrypted and never shown again after saving. An uploaded .env file is not deployed. Only public variables that end up in the frontend bundle anyway, like VITE_ and NEXT_PUBLIC_, are passed into the build automatically.
Step three: export your database and uploaded files
Zeabur backs up databases automatically, but those backups disappear with the platform, and backup download links in the dashboard expire after 15 minutes. Don't treat them as your backup. Export your own copy with the official tool for the same major version as the source database.
For PostgreSQL, run pg_dump per database and export plain SQL. Zeabur's guide warns against restoring a pg_dumpall file as is: its ALTER ROLE ... PASSWORD statements overwrite the new password with the old hash, and its CREATE ROLE and CREATE DATABASE statements clash with what the new environment already created. Inside Zeabur containers, connect to 127.0.0.1 rather than localhost, which has been seen to hang.
pg_dump -h 127.0.0.1 -U <user> -d <dbname> --no-owner --no-acl | gzip > app.sql.gz
mariadb-dump -h 127.0.0.1 -u <user> -p --single-transaction <dbname> | gzip > app.sql.gz
For MySQL or MariaDB, use mysqldump or mariadb-dump. Databases on Vibe Coding Hosting are off until you turn one on: PostgreSQL 16 with pgvector, or MariaDB 11.4, which is MySQL compatible. Connection details arrive as DATABASE_URL and related variables. You can import in two ways: your AI requests a one-time upload URL over MCP and sends a .sql or .sql.gz file of up to 95MB, or you upload up to 32MB from the dashboard. For anything larger, email [email protected] and we'll help. The platform takes a snapshot before every import. Afterwards, compare row counts table by table. Zeabur's guide also recommends actually restoring a backup before cutover: a dump file that exists is not a backup that works.
MongoDB and Redis aren't offered here. If your app stores data in them, it needs rewriting for PostgreSQL or MariaDB, or you'll need a service that provides those databases. If Redis only held a cache or sessions, moving that into the database is usually the simpler fix.
User uploads such as images and attachments usually lived on a Zeabur volume. Here they belong in /data, the persistent folder that survives redeploys. If your app writes to a relative path like uploads, you can map that path to /data without changing code.
Step four: deploy Next.js, Node.js or Python on the new host
Vibe Coding Hosting doesn't run Dockerfiles. It builds with a platform-maintained runtime and detects the project type: package.json means Node.js 22 or 24, requirements.txt means Python 3.12 or 3.13, composer.json means PHP 8.4, and a plain index.html or a Vite frontend is treated as a static site. A leftover Dockerfile is fine because the platform looks at the app itself, but a project with only a Dockerfile and none of those files won't be recognized.
Next.js works with the usual next build and next start, and next start already reads PORT; we've tested Next.js 16. For Express, Fastify and other Node servers, make sure listen uses process.env.PORT on 0.0.0.0. For FastAPI, the start command is uvicorn main:app --host 0.0.0.0 --port $PORT, and Flask or Django under gunicorn work the same way. Django also needs SECRET_KEY and DEBUG=False set as environment variables. Put schema migrations such as npx prisma migrate deploy or python manage.py migrate in the release command; the platform snapshots the database before running it on each deploy.
There are three ways to get the code onto the host. If it's on GitHub, connect the repo on the project page and every push to the chosen branch deploys automatically; a private repo just needs a read-only deploy key. That replaces Zeabur's deploy on push, and the steps are in our GitHub auto-deploy walkthrough. If you use Claude Code, Codex or Gemini CLI, create a token on the project page, paste the setup command into your tool, and ask it to deploy the project to RoamerHost and fix anything the logs complain about. If all you have is files, zip them and upload in the dashboard; leave out node_modules and .venv, since dependencies are reinstalled from your lockfile.
A new version only goes live after it passes a health check. If it fails, the old version keeps serving and you're told what went wrong. The last 3 successful versions are kept for one-click rollback, though rolling back doesn't restore the database. Here's how Zeabur's features map:
| On Zeabur | On Vibe Coding Hosting |
|---|---|
| Deploy on push from GitHub | GitHub auto-deploy, one repo and one branch per project, private repos supported |
| Deployment history, rollback | Deployment logs, last 3 successful versions kept for rollback |
| Automatic database backups | Automatic backups, plus database snapshots you take yourself |
| Volumes | The persistent /data folder |
| Jobs that run on a timer | Schedules: the platform calls a URL in your app, every 5 minutes at most, up to 5 schedules |
*.zeabur.app address | yourname.roamerhost.com with automatic SSL, plus one custom domain |
| Several services, Redis, workers in one project | Not supported: one plan hosts one site |
Step five: point your domain at the new host
Test everything on the default yourname.roamerhost.com address before touching your domain. A day or two before cutover, lower the TTL on your DNS records so both the switch and any rollback take effect faster.
Add the domain on the project page or through your AI. The platform tells you which DNS records to set at your registrar; once they're in place, run the verification and SSL is issued and renewed automatically. Each plan takes one domain: the root domain together with www, or a single subdomain. If you split frontend and backend across app.example.com and api.example.com on Zeabur, you'll need one plan for each, or merge them into a single project, which Next.js handles naturally. If you bought the domain through Zeabur, it can be transferred out from the dashboard; do that before November 30.
New address, new settings elsewhere: update OAuth callback URLs for Google or GitHub sign-in, webhook URLs for payments and LINE, and Supabase's Site URL and Redirect URLs. Avoid hard-coding your address in code; build URLs from the request's Host header, or set an APP_URL variable yourself.
Final checks before you go live
Zeabur's guide puts it plainly: a 200 response and working HTTPS don't mean the migration is done. Before cutover, walk through the main flows on desktop and mobile. Sign up and sign in once, submit a form, trigger a payment or webhook at least once, wait for a schedule to run, and upload an image to confirm it survives a redeploy. Compare database row counts again and make sure email works; send mail through your email provider's HTTPS API.
The order matters: suspend the app on Zeabur, sync data one final time, then change DNS. Suspend rather than delete. If the new environment misbehaves, point DNS back and run npx zeabur@latest service restart to bring the old service back. That fallback only lasts until your Zeabur service expires, and dedicated servers shut down on their own expiry dates, which may come before December 31. Once the new host has been stable for a few days, disconnect Zeabur from your GitHub repo.
Apps that aren't a good fit for Vibe Coding Hosting
It's better to know this before you're halfway through. Vibe Coding Hosting runs websites and web apps that serve HTTP at a URL, one site per plan. You'll want a service that supports these if you need several services in one project, long-running background workers or queues, your own Redis or MongoDB, bots with no web interface such as Discord bots or schedule-only scripts, a Dockerfile or custom image, other languages like Go, Rust or Java, a GPU, or root access and remote command execution.
On the other hand, a Next.js site, an Express or FastAPI API with PostgreSQL, a Laravel project, or a React and Supabase frontend built with Lovable or Bolt are all within scope. If what you actually need is a brand site, a blog or an online store and you'd rather stop dealing with code, look at WP × Roamer AI. For how hosting types divide the work, see shared hosting, VPS and managed hosting compared.
Which plan fits
If you run your own Next.js, Node.js or Python site on Zeabur, already use AI coding tools like Claude Code or Codex, and want someone else looking after the server, with billing in Taiwan dollars, invoices and support in Chinese, choose Vibe Coding Hosting Basic, from NT$299 a month (NT$3,588 billed yearly), and deploy with your own AI over MCP. If someone else wrote the app and you'd like to change it and add features by chatting with an AI after the move, choose Vibe Coding Hosting Pro, averaging NT$999 a month billed yearly. It includes Roamer AI, which does its best with migrated code but can't guarantee results. Full details are on Vibe Coding Hosting plans and pricing, and every new purchase can be fully refunded within 30 days of payment.
If you need more than one site, several services, workers or Redis, look for a service built for that. Short on time? Export your database and the real values of your environment variables first. Those two are the hardest to recover later.
Frequently asked questions
Q: Do I need to change my Next.js code to move it off Zeabur?
Usually not. A typical Next.js project runs on Vibe Coding Hosting with next build and next start, and next start picks up the platform's PORT. What usually needs changing is hard-coded zeabur.app URLs, variables Zeabur injected, and user uploads, which should move to /data.
Q: Can I export all my Zeabur environment variables at once?
Yes, with npx zeabur@latest --json variable list and project export, but the output keeps unexpanded ${VAR} references, and variables shared by other services are listed separately under readonlyVariables. Regenerate every key and password before using them in the new environment.
Q: My database dump is over 95MB. What now?
Compress it to .sql.gz first, which usually shrinks it a lot. If it's still too big, email [email protected] and we'll help you import it. Very large single INSERT statements can also fail, so dumping rows as separate statements is safer.
Q: Can you do the migration for me?
Your own code is deployed by you or your AI. The platform makes sure it runs, and the Basic plan doesn't include changing your code. If a deploy gets stuck, ask Xiaolang, the assistant at the bottom right of the site, or email [email protected] and we'll check the platform side. We can also help when a database is too large to import.
Sources
- Zeabur: Moving off Zeabur migration guide (October 20), migration reference for coding agents
- Zeabur: Migration, refunds and security incident claims (October 22)
- Zeabur: zbpack source code, the tool Zeabur uses to detect and build projects
- RoamerHost: Vibe Coding Hosting supported stacks and limits