Domains and URLs

setup deploys to <worker>.<account-subdomain>.workers.dev and stores that URL as APP_URL.

The workers.dev URL

• The account subdomain. Dashboard → Workers & Pages → Change next to Your subdomain. This is account-wide: every Worker you run moves at once, so the hostname of this instance changes too.
• The Worker name. npm run setup -- --name my-cogsend, or name in wrangler.personal.jsonc, moves this app to my-cogsend.<subdomain>.workers.dev. Renaming deploys a new Worker: the old one keeps its URL, its secrets and its cron trigger until you delete it (and both count against the five-trigger free-plan limit meanwhile).
• Turn the URL off. "workers_dev": false in the config removes it, so only a custom domain (or a route) can reach the app.

A custom domain

The hostname’s zone has to be in the same Cloudflare account. Add it under Workers & Pages → your Worker → Settings → Domains & Routes → Add → Custom Domain (newer dashboards have a Domains tab with the same flow) and enter cogsend.example.com, or an apex like cogsend.com. Cloudflare writes the DNS record and provisions the certificate; the workers.dev URL keeps answering unless you turn it off.

Or keep the routing in config, in wrangler.personal.jsonc so upstream never sees it:

"routes": [{ "pattern": "cogsend.example.com", "custom_domain": true }],
"workers_dev": false

npm run deploy then provisions the domain and drops the workers.dev URL. If the hostname already serves something else, use a route instead of a custom domain — { "pattern": "cogsend.example.com/*", "zone_name": "example.com" } — which needs a proxied DNS record.

After the hostname changes

1. Pin APP_URL to the new origin. Left unset the app follows the host each request arrives on, but a pinned value does not follow a hostname change, and OAuth redirect URIs and signed media URLs are built from whatever it holds.

   node scripts/wrangler.mjs secret put APP_URL

   npm run secrets:put APP_URL uploads it from .dev.vars instead. Re-running setup leaves an existing APP_URL alone, but it will pin the workers.dev URL again if the secret is missing.

2. Re-register the redirect URI in every OAuth app you created — LinkedIn, Threads, X — as https://<new-host>/api/connections/<platform>/callback; a provider whose registered URI no longer matches answers with a redirect error. Mastodon and Bluesky keep working: their tokens are stored, and the redirect URI is only used while connecting.

3. Point an external pinger at the new host. The Worker’s own cron trigger calls the app in-process, so it is unaffected.

4. Confirm with npm run doctor -- --app-url https://<new-host>.

Where to go next

• Configuration — APP_URL, and the rest of what the instance reads.
• OAuth apps — the redirect URIs to register again after a move.
• Cloudflare Access — an extra gate in front of the new hostname.