levelrail-cli import vercel [dir] moves one project in seven printed steps.
It needs only your Levelrail credentials. It never changes DNS and never
deletes or edits anything on Vercel.
Run with --dry-run first: every step is printed and nothing is changed.
The steps
- Read
vercel.json and package.json: framework, build and output, rewrites, redirects, headers and crons. Regions are ignored with a note.
- List the environment variables. Values come from
--env-file or the shell environment and are never printed. NEXT_PUBLIC_ and VITE_ keys are build time, so they are sent with --env on the upload deploy. A file whose name looks like a secret store is refused unless you pass --allow-secret-path.
- Deploy the directory with the same path as
levelrail-cli deploy.
- Attach each
--domain: apex and www are both attached, www redirects to the apex. No DNS records are written.
- Run the domain cutover check and print the exact DNS records to set.
- Translate redirects and headers where supported and list the rest. Each
crons entry becomes an HTTP job on the app (a GET sending Authorization: Bearer from the CRON_SECRET app env var; the value is never printed). Existing jobs with the same name are kept. For Next.js projects, redirects, rewrites and headers are left to next.config; for other projects each rule is listed as not carried over. regions and functions are reported as not applicable on a self-hosted server.
- Print the final checklist.
After the command
- Test the temporary URL of the app.
- Change the DNS records at your registrar. Lower the TTL a day ahead.
- To go back, run
levelrail-cli apps rollback APP or point DNS back at Vercel.
- The Vercel project, its domains, env vars and integrations stay as they were. Remove them yourself once traffic is confirmed.
Not translated: edge middleware, rewrites with has or missing conditions, image optimization settings, and hosted analytics. The command lists each one it finds.Last modified on November 17, 2026