
How to give away a custom domain ending for $10, not $227,000
Technical · 8 min read · Updated 2026-08-31
Short answer: "I want my own domain ending" almost never means someone actually wants to run a top-level domain. What they want is a short, ownable, brand-shaped name attached to something. ICANN will sell you a real TLD, but the 2026 new-gTLD round charged USD 227,000 per application just to be evaluated, plus ongoing registry operation obligations on top of that. The working alternative costs about $10: register one domain whose name already reads like an ending, point one wildcard DNS record at an app, and give away subdomains through a public database instead of a DNS panel. runs-on.dev is a live example, one we built at Advance Labs. It isn't a TLD.
## what_people_mean
What people actually mean by "my own domain ending"
There are three real ways to get something that behaves like a custom domain ending, and they're not interchangeable. The first is to buy a normal domain and stop there. yourname.com is already "your own ending" in every practical sense; nobody else can register under it, and you decide what goes at every subdomain. Most people asking for a custom TLD would be happy with this alone.
The second is to run an opt-in alternative DNS root, the kind of project that mints endings like .crypto or .zil outside ICANN's system. These resolve for whoever installs the matching resolver or browser extension, and for nobody else. They look like a new TLD in a screenshot and behave like nothing on the actual internet unless the visitor already opted in.
The third is the real thing: apply to ICANN and get an entry in the root zone that every standard resolver on earth already understands. That's the only option that produces an ending like .com or .dev everyone's browser resolves by default with no extra software. It's also the expensive one.
## icann_reality
The ICANN reality
ICANN opened a new round of generic top-level domain applications, accepted from April 30 to August 12, 2026. The evaluation fee was USD 227,000 per application, non-refundable in whole once evaluation begins, and that number covers evaluation only. Get approved and you're also on the hook for registry operation. That means the technical infrastructure to actually run it, staying in compliance, and fees that don't stop once the TLD is live.
The number that matters
USD 227,000 per application, just to be evaluated, in the round that ran April 30 to August 12, 2026. That's before you've sold a single subdomain or stood up a single line of registry infrastructure.
For almost everyone who says "I want my own TLD," that fee is the whole conversation. It's not a startup cost you grow into. It's a gate that assumes you're a registry operator, a large brand protecting a trademark string, or a country administering a ccTLD, not someone who just wants a clean namespace for a community, a hobby project, or a product.
## the_approach
The approach that actually works
Register one ordinary domain whose name reads like an ending, then give away subdomains under it. That's it. is-a.dev did this for developer portfolios. js.org did it for JavaScript projects. eu.org has been doing it since long before either. None of these are TLDs. Each is one domain, with a very long list of people who each control one subdomain of it.
The trick is where the list of subdomains lives. Instead of a DNS panel only the domain owner can touch, the list lives as a public GitHub repository. Claiming a name becomes a pull request: add a file, get it merged, and your subdomain exists. Nobody needs registrar access, and the whole history of who claimed what sits in a public git log.
## wildcard_mechanism
The wildcard mechanism, explained properly
This only works because of one DNS record: a wildcard, written as *.example.com, pointed at a single application. Without it, every new subdomain needs its own DNS entry added by hand, which is exactly the bottleneck a registrar runs into. With it, anything.example.com, whatever anything is, resolves to the same app, which then reads the incoming hostname and decides what to serve.
TLS follows the same trick. A wildcard certificate for *.example.com covers every subdomain automatically, present or not yet claimed, so a brand-new name gets valid HTTPS the moment DNS propagates, with no per-subdomain certificate request. Put those two together and claiming a name stops being a DNS operation at all. It becomes a database write: add a row (or in this design, merge a file), and the wildcard record and wildcard certificate already in place do the rest.
One wildcard A/CNAME record plus one wildcard TLS certificate is enough to serve an unlimited number of subdomains that don't exist yet. Every future claim becomes a database write, not a DNS change.
## why_dot_dev
Why .dev specifically
The .dev TLD is on the HSTS preload list, the list browsers ship with that forces HTTPS before the first request ever goes out, with no trust-on-first-use plaintext connection possible. Because the preload entry covers the whole TLD, every subdomain under it inherits the same guarantee automatically. A visitor can't accidentally load name.example.dev over plain HTTP, even on the very first visit, even if a link or a typo pointed them at http://.
For a registry handing subdomains to people you haven't vetted, that's a security property, not a branding choice. It closes off an entire class of downgrade and interception attacks on every name in the registry, without asking each subdomain owner to configure anything.
## underestimated_parts
The parts people underestimate
Running the technical side is the easy part. What actually takes ongoing attention comes from being the registrant of record for every name someone else claims:
- Abuse and impersonation: Someone will try to claim paypal, google, or a name that shadows a real brand or a real person. You need a reservation list and a review step before a claim merges, not just a merge-on-CI-pass pipeline.
- Name grammar and homoglyphs: You have to decide what characters are even legal in a claimed name, and reject names built from lookalike Unicode characters designed to impersonate an existing, legitimate claim.
- Reserved names: www, api, mail, admin, and anything that could be mistaken for registry infrastructure itself needs to be off the table before the first outside claim, not patched in after someone already takes it.
- You answer for what every subdomain serves: Legally and practically, you're the domain owner. If a claimed subdomain starts serving phishing pages or malware, the abuse reports, hosting-provider takedowns, and browser blocklist entries land on your domain, not on the individual claimant's.
## working_example
A working example
runs-on.dev is our version of this pattern at Advance Labs, live today. Anyone can claim a name.runs-on.dev subdomain with a GitHub account in seconds and get a profile card out of it. Pointing the name at their own hosting is a pull request against a public registry repository. The whole cost of standing it up was about $10, one domain, one year, against the $227,000 ICANN charges just to evaluate a real TLD application.
It's not a TLD, and we don't claim it is. It's one .dev domain with a wildcard record and a public git-based registry sitting behind it, the same shape as is-a.dev, js.org, and eu.org before it. See it at
## faq
Is runs-on.dev a top-level domain?
No. runs-on.dev is one registered domain (a .dev subdomain registry) that gives away subdomains under itself. It is not a new gTLD, was never submitted to ICANN, and does not appear in the DNS root zone as its own ending. Claiming a name gets you name.runs-on.dev, a subdomain, not a standalone TLD.
How much does ICANN charge for a new top-level domain?
In the 2026 new-gTLD application round, which ran April 30 to August 12, 2026, ICANN's evaluation fee was USD 227,000 per application. That covers evaluation only; a successful applicant also takes on ongoing registry operation obligations and fees.
How does a subdomain registry give out HTTPS for names that don't exist yet?
A wildcard TLS certificate for *.example.com covers every possible subdomain in advance, so a name gets valid HTTPS the moment it's claimed and DNS propagates. No per-subdomain certificate request is needed.
Why use a .dev domain for a registry like this instead of .com?
.dev is on the browser HSTS preload list, so every subdomain under it is forced to HTTPS from the first request, even before any certificate has been seen by that browser. For a registry serving pages for people you haven't vetted, that closes off a class of downgrade attacks by default.
What's the actual mechanism that makes claiming a name fast?
One wildcard DNS record (*.example.com) pointed at one application, combined with a wildcard TLS certificate. Both are set up once. After that, claiming a subdomain is a database write, in runs-on.dev's case a merged pull request against a public GitHub repository, not a DNS change.
## sources
## related_guides
## claim_a_name
runs-on.dev is free and live. Claim name.runs-on.dev with a GitHub account, then point it at your own hosting with a pull request against the public registry.
Open runs-on.dev