A WIP Go rewrite and extension of TLDHunt. A CLI to check domain name availability.

$ go install github.com/miku/tldhunter@latest

$ tldhunter -h
Usage: tldhunter -k <keyword|domain> [-e <tld> | -E <tld-file>] [-x] [-v] [--update-tld]
Without -e or -E, the built-in TLD list (1438 entries) is used,
unless the keyword already ends in a known TLD, which checks just that domain.
Results are cached in /Users/tir/.cache/tldhunter for 24h0m0s (1h0m0s if available; -ttl 0 to disable).
Example: ./tldhunter -k linuxsec
       : ./tldhunter -k delta.sh
       : ./tldhunter -k linuxsec -E tlds.txt
       : ./tldhunter -k linuxsec -e .dev -v
       : ./tldhunter --update-tld
       : ./tldhunter --clear-cache

-v (or --verbose) dumps every whois and RDAP response to stderr, prefixed with [debug], along with failed attempts and cache hits. Results still go to stdout, so tldhunter -k linuxsec -v 2>debug.log keeps the two apart.

Everything lives in a single file, tldhunter.go, and uses only the standard library -- no whois(1), no curl(1).

• Keyword times TLD list. The keyword is combined with every TLD in tlds.txt, which is embedded into the binary at build time. A keyword that already ends in a known TLD (delta.sh) is read as a single domain instead of a scan; -e checks one TLD, -E reads a list from a file.
• Registry lookup, per TLD. Each TLD is resolved to its registry endpoint by asking whois.iana.org on TCP/43 (RFC 3912). If IANA publishes no server, the conventional whois.nic.<tld> host is tried when it resolves.
• RDAP fallback. A growing number of registries (.dev and the rest of Google's are the standard example) answer only over RDAP. Those TLDs are looked up over HTTPS (RFC 7480/9082) using endpoints from IANA's bootstrap registry (RFC 9224). Handoff also happens mid-run, when a published whois host turns out to disown the TLD or has closed port 43.
• Available first, then registered. A whois response is matched line by line against explicit "no match"/"not found"/"status: free" patterns first, and only then against registration evidence (name servers, creation date, registrar, expiry). The shell original did the reverse, so any response it did not understand read as an available domain -- a registered .de, whose whois says only Domain: x.de / Status: connect, was reported free. Anything matching neither set is printed as unknown rather than guessed at. Over RDAP there is nothing to match: a 404 is the answer.
• Grouped concurrency. Domains are grouped by endpoint and each group is capped at -perhost concurrent queries (-j bounds the total). This matters because hundreds of gTLDs share one host -- a single Identity Digital RDAP endpoint serves 451 TLDs in the list -- so a flat worker pool would aim most of its concurrency at one machine and get throttled. Refusals and resets are retried with jittered exponential backoff, honouring Retry-After.
• On-disk cache. Verdicts, TLD-to-endpoint mappings, and the RDAP bootstrap registry are cached under $XDG_CACHE_HOME/tldhunter as one small JSON file per entry, written atomically. Only definitive answers are stored, never a network failure. Taken domains keep for 24h (-ttl), available ones for 1h (-ttl-avail), since those are the ones that go stale in the direction that costs something. -ttl 0 disables the cache, --clear-cache removes it.

If you have a better signature or detection method, please feel free to submit a pull request.

For the default Top Level Domain list (tlds.txt), we use data from https://data.iana.org. You can update this list directly using the --update-tld flag, which fetches the latest TLDs from IANA and saves them to tlds.txt.

You can also use a custom TLD list, but ensure it is formatted like this:

.aero
.asia
.biz
.cat
.com
.coop
.info
.int
.jobs
.mobi